add a new vignette on collinearity
new data set base_pub, base on publication data from the Microsoft Academic Graph, used to illustrate the new vignette
more notes and messages better fit the user screen
default for the argument collin.tol increased from 1e-10 to 1e-9.
fix bug when several scalars to be evaluated from the calling environment are included in the formula of the regression
fix bug when: using IV, interpolating the LHS variable, requesting multiple LHS, using a RHS formula as interpolated variable. Reported by @sumtxt, #522
fix bug lean = TRUE: remove a potentially large a copy of environments when an estimation was called within a function. Thanks to @aaronrudkin, #514
fix bug in etable: some arguments globally set were wrongly reset when calling the function. Reported by @aaronrudkin, @etiennebacher kindly provided a reprex, #517
fix documentation bugs. Reported by @cseveren
stringmagic version >= 1.1.2 to fix several major bugs affecting R versions <= 4.1.2fix bug leading to CRAN error (in IV with fixed-effects and no exogenous variable)
fix display bug in errors regarding lead/lag problems in fixest estimations.
fix bug when arguments lead or lag in functions l/f were given explicitly in fixest estimations. Reported by @MaelAstruc, #499
remove the diplay of TRUE when creating lags/leads in data.table.
fix bug when creating lead/lags in data.tables and the name of the data set was equal to the name of a function. Reported by @kmfrick, #500
fix several bugs affecting R version 3.5.0
fix bug in etable when variables appeared in the statistics and those variables contained invalid Latex characters. PR by @MaelAstruc, #508
fix formatting for coefficients/statistics equal to 0. Reported by @MaelAstruc, #504
the demeaning algorithm has been reworked! The code has been condensed and four new parameters have been introduced to control the details of how the algorithm works. The new default values should lead to quicker convergence in general (for difficult cases). The new function demeaning_algo gives the user fine control over the internal parameters of the algorithm.
all estimation functions gain the argument data.save. If TRUE, the data set used for the estimation is saved in the returned object. This ensures the consistency of post-processing (like fit statistics, predict, update, etc) even if the original data has been modified in the meantime. Suggestion by Vincent Arel-Bundock, #340
new function fixest_data to access the original data set used at estimation-time. Suggestion by Kyle Butts, #465
new method df.residual.fixest. Suggestion by @rferrali, #455
now the vcov method inherits the small sample correction and type of VCOV used in the original estimation. Thanks to @mgoplerud, #356
new function n_models to identify the length of a dimension of a multiple estimation. This facilitates the manipulation of multiple estimations programmatically:
base = setNames(iris, c("y", "x1", "x2", "x3", "species"))
est = feols(y ~ csw(x1, x2, x3), base, fsplit = ~species)
# We can obtain the unique number of RHSs/samples with `n_models`
n_models(est, rhs = TRUE)
#> [1] 3
n_models(est, sample = TRUE)
#> [1] 4
all_tables = list()
for(i in 1:n_models(est, sample = TRUE)){
all_tables[[i]] = etable(est[sample = i])
}
do.call(rbind, all_tables)
# ... the output is too long to be displayed here
fix bug when the covariance matrix was very ill-defined. Thanks to Gianluca Russo.
pass correctly nthreads to summary when the VCOV is provided at estimation time. Thanks to @arcruz0, #379.
fix bug in the function demean when a formula is passed as argument and the data was in the form of a data.table. Reported by Patrick Baylis, with help from Kyle Butts, #433
fix bug when an offset was used in the context of multiple outcomes. Reported by @etiennebacher, #405
fix bug in the collinearity function. Reported by @grlju, #412
fix bug which allowed the estimation of models with variables from the environment while it shouldn't
fix bug in the demeaning code when the data set in input is of length 1. Reported by @da-zar with help of @etiennebacher, #283
fix bug in predict when the offset was placed in a formula and not passed as a regular argument. Reported by @jl-flores, #309
fix bug in TSLS estimations when the formula contained multiple LHS to be expanded with the dot square bracket operator. Reported by @svraka, #395
fix bug leading to the absence of warning when the convergence of the demeaning algorithm failed. Thanks to @ja-ortiz-uniandes, #323
fix bug in IV estimation when all exogenous variables were removed because of collinearity with the fixed-effects. Thanks to @Oravishayrizi, #371
fix bug in update.fixest when no FEs were used in the initial estimation and one wanted to add FEs. Reported by @hughjonesd, #475
fix bug update.fixest: single estimations extracted from multiple estimations can now be updated individually. Note however that if split was used in the main estimation, the subsample does not carry trhough (this limitation will be fixed later).
functions vcov gains the argument vcov_fix to monitor whether to fix the covariance matrix with an eigenvalue decomposition (previously this was always turned on).
in estimations, the algorithm now allows to use scalars from the environment which are non numeric. Thanks to @kennchua and @kylebutts, #426
improve error messages in iplot.
improve speed of the function fixef.fixest in the presence of varying slopes. PR of @etiennebacher, #390
the subsetting of multiple estimations does not error any more when the index used was not relevant in the multiple estimation (provided its value is equal to 1)
new method update.fixest_multi.
fix typos and display issues thanks to: @statzhero, @eyayaw, @etiennebacher, @raffaem.
add missing documentation of mvsw, thanks to @grantmcdermott, #425
fix bug "t value" displaying in lieu of "z value". Thanks to issue on easystats/parameters #892. Very ancient bug!
fix bug in coefplot: now the confidence intervals use the Student t when appropriate (before that, only the Normal law was used). Thanks to Grant McDermott, #409
fix broken links in documentation (to the fixest paper)
fix bad documentation markup
adapt to CRAN policy to remove multithreading when checking the package
binfix bug in function coef() leading to methods to throw errors in R devel. Thanks to @vincentarelbundock for reporting (#291).
fix bug in the predict method when applied to objects estimated with feNmlm. Thanks again to @vincentarelbundock for reporting (#292)!
fix missing variable names in the VCOV matrix of feNmlm models. Thanks (yet again!) to @vincentarelbundock for reporting (#293). Comme on dit : jamais deux sans trois !
fix display bug in cluster names in etable.
fix bug in IV estimations with no exogenous variable and no fixed-effect (thanks to Kyle Butts, #296).
fix bug panel vs panel.id behaving differently in terms of default type of VCOV when the estimation did not contained lags.
fix bug in confint.fixest when only one variable was estimated (thanks to @joachim-gassen, #296).
fix several bugs in predict when using i(), in particular when used in combination with a factor or poly() (reported by @rfbressan, #301).
fix bug in etable relating to ampersands not being correctly escaped.
fix bug in sunab when the time variable is exactly named t (reported by Florian Hollenbach, #330).
fix bug in feols.fit when vcov was supplied and the estimation did not contain fixed-effects (reported by @grlju, #341).
fix bug in sample_df when the name of the variable was too long.
fix rare bug regarding an error message when a missing variable did exist as a function in the environment.
fix bug preventing the use of binning with formulas reported by @tlcaputi, #359).
fix various errors in the documentation (thanks to Ed Rubin and others!).
fix bug in warning message in peculiar case of divergence in GLM (reported by @pachadotdev, #315).
fix bug preventing the use of the global data set in wrapper functions (fepois, fenegbin, etc). Reported by @turbanisch, #343.
fix bug preventing the use of split in non-GLM, non-OLS estimations (reported by @bberger94, #333).
new internal algorithm leading to an object very much like a plain list, much easier to interact with.
new function models to extract the matrix of reporting which model has been estimated.
base = setNames(iris, c("y", "x1", "x2", "x3", "species"))
mult_est = feols(y ~ csw(x.[,1:3]), base)
models(mult_est)
#> id rhs
#> 1 1 x1
#> 2 2 x1 + x2
#> 3 3 x1 + x2 + x3
in multiple estimations: all warnings are turned to notes and all notes are delayed and stacked.
coef.fixest_multi: Now reports the model of each estimation in the first columns. Also gains the arguments collin, long (to display the results in a long format) and na.rm.
coef(mult_est)
#> id rhs (Intercept) x1 x2 x3
#> 1 1 x1 6.526223 -0.2233611 NA NA
#> 2 2 x1 + x2 2.249140 0.5955247 0.471920 NA
#> 3 3 x1 + x2 + x3 1.855997 0.6508372 0.709132 -0.5564827
# Now in long format
coef(mult_est, long = TRUE)
#> id rhs coefficient estimate
#> 1 1 x1 (Intercept) 6.5262226
#> 2 1 x1 x1 -0.2233611
#> 5 2 x1 + x2 (Intercept) 2.2491402
#> 6 2 x1 + x2 x1 0.5955247
#> 7 2 x1 + x2 x2 0.4719200
#> 9 3 x1 + x2 + x3 (Intercept) 1.8559975
#> 10 3 x1 + x2 + x3 x1 0.6508372
#> 11 3 x1 + x2 + x3 x2 0.7091320
#> 12 3 x1 + x2 + x3 x3 -0.5564827
coeftable.fixest_multi, se.fixest_multi, tstat.fixest_multi, pvalue.fixest_multi to easily extract the results from multiple estimations.coeftable(mult_est)
#> id rhs coefficient Estimate Std. Error t value Pr(>|t|)
#> 1 1 x1 (Intercept) 6.5262226 0.47889634 13.627631 6.469702e-28
#> 2 1 x1 x1 -0.2233611 0.15508093 -1.440287 1.518983e-01
#> 3 2 x1 + x2 (Intercept) 2.2491402 0.24796963 9.070224 7.038510e-16
#> 4 2 x1 + x2 x1 0.5955247 0.06932816 8.589940 1.163254e-14
#> 5 2 x1 + x2 x2 0.4719200 0.01711768 27.569160 5.847914e-60
#> 6 3 x1 + x2 + x3 (Intercept) 1.8559975 0.25077711 7.400984 9.853855e-12
#> 7 3 x1 + x2 + x3 x1 0.6508372 0.06664739 9.765380 1.199846e-17
#> 8 3 x1 + x2 + x3 x2 0.7091320 0.05671929 12.502483 7.656980e-25
#> 9 3 x1 + x2 + x3 x3 -0.5564827 0.12754795 -4.362929 2.412876e-05
confint.fixest_multi to extract the confidence intervals of multiple estimations.new argument add to facilitate adding elements to the formula.
new argument frame to tell where to fetch the values of the variables expanded with the dot square bracket operator.
empty strings or empty elements expanded with .[] are now set to be equal to 1 (the neutral element in formulas):
x = ""
xpd(y ~ .[x] + .[NULL])
#> y ~ 1 + 1
!:xpd(am ~ ..("!^am"), data = mtcars)
#> am ~ mpg + cyl + disp + hp + drat + wt + qsec + vs + gear + carb
base = setNames(iris, c("y", "x1", "x2", "x3", "species"))
xpd(y ~ x.., data = base)
#> y ~ x1 + x2 + x3
feols(y ~ x.., base)
#> OLS estimation, Dep. Var.: y
#> Observations: 150
#> Standard-errors: IID
#> Estimate Std. Error t value Pr(>|t|)
#> (Intercept) 1.855997 0.250777 7.40098 9.8539e-12 ***
#> x1 0.650837 0.066647 9.76538 < 2.2e-16 ***
#> x2 0.709132 0.056719 12.50248 < 2.2e-16 ***
#> x3 -0.556483 0.127548 -4.36293 2.4129e-05 ***
#> ---
#> Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#> RMSE: 0.310327 Adj. R2: 0.855706
xpd in non-fixest functions, the algorithm tries to guess the data so that calls to ..("regex") or auto-completion can be used seamlessly.lm(xpd(y ~ x..), base)
#> Call:
#> lm(formula = xpd(y ~ x..), data = base)
#>
#> Coefficients:
#> (Intercept) x1 x2 x3
#> 1.8560 0.6508 0.7091 -0.5565
xpd also expands one-sided formulas:x_all = ~sepal + petal
xpd(color ~ .[x_all])
#> color ~ sepal + petal
fitstat, the formula is now automatically expanded with xpd. This means that you can set fit statistics macro which can be summoned from etable. Useful to set default fit statistics for: IVs, GLMs, etc.base = setNames(iris, c("y", "x1", "x2", "x3", "species"))
est = feols(y ~ csw(x.[,1:3]), base)
# setting the macro
setFixest_fml(..fit_ols = ~ n + ar2 + my)
# summoning it
etable(est, fitstat = ~..fit_ols)
#> est.1 est.2 est.3
#> Dependent Var.: y y y
#>
#> Constant 6.526*** (0.4789) 2.249*** (0.2480) 1.856*** (0.2508)
#> x1 -0.2234 (0.1551) 0.5955*** (0.0693) 0.6508*** (0.0667)
#> x2 0.4719*** (0.0171) 0.7091*** (0.0567)
#> x3 -0.5565*** (0.1275)
#> _______________ _________________ __________________ ___________________
#> S.E. type IID IID IID
#> Observations 150 150 150
#> Adj. R2 0.00716 0.83800 0.85571
#> Dep. Var. mean 5.8433 5.8433 5.8433
now there is support for models with no coefficient (only fixed-effects).
the application of markdown markup is now more robust and can also be escaped with a backslash. The escaping has been ported to c++.
list. If TRUE, then the result is returned in a list form. Useful in Rmarkdown documents for quick reference to specific values.est = feols(mpg ~ cyl + drat + wt, mtcars)
ct = coeftable(est, list = TRUE)
ct$constant$coef
#> Estimate
#> 39.76766
ct$wt$se
#> Std. Error
#> 0.8293065
split and fsplit gain the %keep% and %drop% operators which allow to split the sample only on a subset of elements. All estimations also gain the arguments split.keep and split.drop which do the same thing as the previous operators.base = setNames(iris, c("y", "x1", "x2", "x3", "species"))
est = feols(y ~ x.[1:3], base, fsplit = ~species %keep% c("set", "vers"))
etable(est)
#> model 1 model 2 model 3
#> Sample (species) Full sample setosa versicolor
#> Dependent Var.: y y y
#>
#> (Intercept) 1.856*** (0.2508) 2.352*** (0.3929) 1.896*** (0.5071)
#> x1 0.6508*** (0.0667) 0.6548*** (0.0925) 0.3869. (0.2045)
#> x2 0.7091*** (0.0567) 0.2376 (0.2080) 0.9083*** (0.1654)
#> x3 -0.5565*** (0.1275) 0.2521 (0.3469) -0.6792 (0.4354)
#> ________________ ___________________ __________________ __________________
#> S.E. type IID IID IID
#> Observations 150 50 50
#> R2 0.85861 0.57514 0.60503
#> Adj. R2 0.85571 0.54743 0.57927
as.dict:x = "
# Main vars
mpg: Miles per gallon
hp: Horsepower
# Categorical variables
cyl: Number of cylinders; vs: Engine"
as.dict(x)
#> mpg hp cyl vs
#> "Miles per gallon" "Horsepower" "Number of cylinders" "Engine"
# setFixest_dict works directly with x
setFixest_dict(x)
setFixest_dict: i) now the dictionary only grows, ii) you can define variables directly in the arguments of setFixest_dict, iii) as.dict is applied to the dictionary if relevant, iv) there's a new argument reset.new function degrees_freedom_iid which is a more user-friendly version of degrees_freedom.
new function fdim to print the dimension of a data set in an user-readable way.
remove warnings when a binomial family is used with weights in feglm.
add the arguments y, X, weights, endo, inst to the function est_env to make it more user-friendly.
fix documentation typos (thanks to Caleb Kwon).
etable now returns a data.frame whose first column is the variables names (before this was contained in the row names).
fix environment problems when lean = TRUE, leading to large objects when saved on disk.
print.fixest now displays the information on the sample/subset/offset/weights.
fixef) when there are 3+ fixed-effects. This bug led to, in some specific circumstances, wrong values for the fixed-effects coefficients. Thanks a lot to @pachadotdev (#286) for finding this out!fix bug in confint when sunab was used (thanks to Sarah Hofmann).
fix an important "documentation bug" on the Sun and Abraham method (thanks to Kyle Butts, #287).
fix bugs regarding view/markdown features of etable.
added compatibility with car::deltaMethod following Grant McDermott's suggestion.
new function lag_fml which is an alias to lag.formula. The latter being easily stomped by other function names from other packages.
fix bug linked to the proper identification of estimations with only fixed-effects.
remove the use of anyNA.data.frame leading to a dependency to R 3.6.3 (reported by @MichaelChirico, #261).
fix bug in stepwise estimations when two (stepwised) explanatory variables have exactly the same NAs values.
fix display bug regarding factors in etable when dict was present.
fix bug tablefoot.value not working any more (reported by @resulumit, #224).
fix possible environment problem when estimating non linear functions outside of the global environment.
fix bug in the stepwise functions sw and csw when they contained only one variable.
fix bug in etable preventing automatic headers to be displayed.
fix bug in n_unik preventing the auto completion of variable names.
fix bug in fitstat for the KPR statistic (reported by @etiennebacher, #161).
fix bug in i when two factor variables were interacted and one specific value of one variable was to be set as a reference.
fix bug in model.matrix when no variable was used in the estimation (reported by @kylebutts, #229).
model.matrix now returns the variables in the same order as in the estimation -- a discrepancy could happen in stepwise estimations with interactions in which the interactions were put before fixed covariates (related to @sergiu-burlacu, #231).
fix bugs in feglm.fit prevented the VCOV to be computed (reported by @etiennebacher and @edrubin, #237)
fix bug in predict with variables created with i() leading to a prediction even for values not included in the original estimation (reported by @vincentarelbundock, #235).
fix bug in multiple estimations when the data contains weights and there are missing values in the y's or X's (reported by @sahilchinoy, #263).
fix bug multiverse stepwise when the estimation contains fixed-effects or IVs (reported by @resulumit, #260).
fix bug in the startup message trigger.
increase the robustness of the code leading to the startup message (reported by @flycattt, #262).
improve the robustness of the algorithm parsing the fixed-effects (linked to issue, #253).
fix minor bug in the Cragg-Donald statistic.
fix peculiar problem on load when directories names end with ".R" (thanks to @kyleam, #271).
remove remaining large items from GLM estimations with lean = TRUE.
fix bug in removing the singletons from several fixed-effects (reported by @johannesbubeck, #244).
in rep.fixest: replace argument cluster with argument vcov to enable the use of any VCOV (related to, #258 by @ShunsukeMatsuno).
fix bug in predict, which automatically discarded NA values (reported by @ColinTB, #273).
new argument view to display the latex table in the viewer pane (suggestion by Or Avishay-Rizi, #227). You need to a) have a working distribution of pdflatex, imagemagick and ghostscript, or b) have the R packages pdftools and tinytex installed, for this feature to work.
new argument view.cache in setFixest_etable: whether to cache the PNGs generated.
new argument export to export the Latex table in PNG to a file.
new (experimental) argument markdown: Latex tables can be automatically integrated in the non-Latex markdown document in PNG format.
new argument div.class. Linked to the markdown argument. In Rmarkdown documents, the table in PNG format is embedded in a <div> container. The class of the div is div.class, which is by default "etable".
new argument tpt to nest the table in a threeparttable environment. Notes are then nested into the tablenotes environment.
in style.tex: new argument notes.tpt.intro to insert code right after the tablenotes environment and before any note (useful to set the font size of notes globally for instance).
new argument arraystretch to set the height of the table rows.
new argument fontsize which applies Latex font sizes to the table.
new argument adjustbox: adjustbox = TRUE nests the tabular into an adjustbox environment with width = \\textwidth, center as default option. Use adjustbox = x with x a number giving the text-width. Use adjustbox = "x th" with x a number giving the text-height. Finally you can use a character string, as in adjustbox = "my options", that will be passed verbatim as a an adjustbox option.
new argument highlight to highlight the coefficients with a frame or by changing the row/cell color.
new argument coef.style to apply an arbitrary style to one or several coefficients.
in style.tex: new argument rules_width to easily set the width of the booktabs rules.
in style.tex: new argument caption.after to insert code right after the caption.
in style.tex: new argument no_border to remove the borders on the sides of the table.
the quality of the tex output has been substantially improved.
signif.code now replaces signifCode (retro compatibility ensured).
signifCode is removed from setFixest_etable, and signif.code is added to both style.tex and style.df so that each style can have its own significance code defined globally.
the object returned by etable are now of class etable_tex (when tex = TRUE) or etable_df, both types having their own printing method.
the significance codes are now displayed under the table when the output is a data.frame.
in headers/extralines: cmidrule does not show up for empty column names any more.
new markup: markdown-style markup (e.g. **text**) can be used to put text in italic/bold in almost anything in the table.
notes can be set in the dictionary: useful for notes (like source for example) that gets repeated across tables.
line.top and line.bottom now admit the values simple and double. The argument line.bottom now affects the "effective" end of table, irrespective of the value of tablefoot. This is more in line with intuition.
improve the use of tabularx.
automatic support makecell: any new lines found in names within the table will be translated with makecell. For example: "The \n long \n varname" is automatically translated into \makecell{The \\ long \\ varname}.
completely new function dsb() to manipulate strings. Applies many low level string operations very easily. The syntax may be a bit disturbing at first, but, unlike French grammar, there's some logic behind!
there are over 30 basic string operations available! Do complex string manipulations in a single call!
# At first sight, it's impossible to understand what's going on.
# But I assure you, it's pretty logical!
# Type dsb("--help") to get some help.
dollar = 6
reason = "glory"
dsb("Why do you develop packages? For .[`dollar`*c!$]?",
"For money? No... for .[U,''s, c?reason]!", sep = "\n")
#> Why do you develop packages? For $$$$$$?
#> For money? No... for G L O R Y!
dsb when the calls are nested:xpd(~ sw(.[, "disp:.[/mpg, cyl]"]))
#> ~sw(disp:mpg, disp:cyl)
only.coef in all estimation. If TRUE, then only the estimated coefficients are returned, which can be useful for MC experiments.est_env to estimate a model from a fixest environment. Mostly useful to cut overheads in simulations.# First we get the environment (the estimation is not performed!)
env = feols(mpg ~ disp + drat, mtcars, only.env = TRUE)
# Then we estimate: we get the reult from feols(mpg ~ disp + drat, mtcars)
est_env(env)
#> OLS estimation, Dep. Var.: mpg
#> Observations: 32
#> Standard-errors: IID
#> Estimate Std. Error t value Pr(>|t|)
#> (Intercept) 21.844880 6.747971 3.23725 3.0167e-03 **
#> disp -0.035694 0.006653 -5.36535 9.1914e-06 ***
#> drat 1.802027 1.542091 1.16856 2.5210e-01
#> ---
#> Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#> RMSE: 3.07661 Adj. R2: 0.712458
# Why doing that? You can modify the env w/t incurring overheads
assign("weights.value", mtcars$wt, env)
# New estimation with weights
est_env(env)
#> OLS estimation, Dep. Var.: mpg
#> Observations: 32
#> Standard-errors: IID
#> Estimate Std. Error t value Pr(>|t|)
#> (Intercept) 21.967576 6.320006 3.47588 1.6241e-03 **
#> disp -0.032922 0.005884 -5.59478 4.8664e-06 ***
#> drat 1.505517 1.470671 1.02369 3.1444e-01
#> ---
#> Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#> RMSE: 5.08781 Adj. R2: 0.709392
ref which allows to re-factor variables on-the-fly. This function always returns a factor and relocates the values given in the argument as the first factor levels. It also allows to bin values, similarly to the function bin:# We want to place 5 in the first place
ref(1:5, 5)
#> [1] 1 2 3 4 5
#> Levels: 5 1 2 3 4
# You can also bin at the same time
ref(1:5, .("4:5" = 4:5))
#> [1] 1 2 3 4:5 4:5
#> Levels: 4:5 1 2 3
bin: cut:: now ignores white spaces, so that cut:: q1 ] q3 [ works appropriately.
speed of stepwise estimations (using sw [not csw]) has been improved.
recursive formula macro definitions are allowed (feature request by @turbanisch, #234).
the startup message does not pop in Rmarkdown documents any more.
function sample_df gains the argument previous which recovers the previous draw.
remove new R native piping test |> which led to errors in R < 4.1.0 despite conditional testing.
fix bug in etable headers when one wants to include several lines and the first line contains only one element repeated across columns.
fix bugs in predict: a) when variables are created with functions of the data, and b) when the new data contains single level factors (relates to issues #200 and #180 by @steffengreup and @IsadoraBM).
fix bug in etable non-clustered standard errors not displaying properly in footers.
fix bug in etable regarding the escaping of fixef_sizes (reported by Apoorva Lal, #201).
fix bug introduced in 0.10.0 preventing the estimation of IV models with interacted fixed-effects (reported by @etiennebacher, #203).
fix bug in IV estimations when: a) no exogenous variables were present AND the IV part contained at lags; and b) the endogenous variables contained at least two lags. Reported by Robbie Minton.
fix bug in the .fit methods when the argument vcov wasn't NULL.
fix bug in summary.fixest_multi: when the variance was NA and internal bug could pop in some circumstances.
fix bug plot.fixef not working for fepois (reported by @statzhero, #213).
fix error message when the (wrong) argument X is used in feols.
.[,stuff], to separate variables with commas (instead of separating them with additions):lhs_vars = c("var1", "var2")
xpd(c(.[,lhs_vars]) ~ csw(x.[,1:3]))
#> c(var1, var2) ~ csw(x1, x2, x3)
new function dsb: applies the dot square bracket operator to character strings.
in the function dsb, you can add a string literal in first or last position in .[] to "collapse" the character string in question. The way the collapse is performed depends on the position:
name = c("Juliet", "Romeo")
# default behavior => vector
dsb("hello .[name], what's up?")
#> [1] "hello Juliet, what's up?" "hello Romeo, what's up?"
# string literal in first position
dsb("hello .[' and ', name], what's up?")
#> [1] "hello Juliet and Romeo, what's up?"
# string literal in last position
dsb("hello .[name, ' and '], what's up?")
#> [1] "hello Juliet and hello Romeo, what's up?"
bin: numeric vectors can be 'cut' with the new special value 'cut::q3]p90]', check it out!data(iris)
plen = iris$Petal.Length
# 3 parts of (roughly) equal size
table(bin(plen, "cut::3"))
#>
#> [1.0; 1.9] [3.0; 4.9] [5.0; 6.9]
#> 50 54 46
# Three custom bins
table(bin(plen, "cut::2]5]"))
#>
#> [1.0; 1.9] [3.0; 5.0] [5.1; 6.9]
#> 50 58 42
# .. same, excluding 5 in the 2nd bin
table(bin(plen, "cut::2]5["))
#>
#> [1.0; 1.9] [3.0; 4.9] [5.0; 6.9]
#> 50 54 46
# Using quartiles
table(bin(plen, "cut::q1]q2]q3]"))
#>
#> [1.0; 1.6] [1.7; 4.3] [4.4; 5.1] [5.2; 6.9]
#> 44 31 41 34
# Using percentiles
table(bin(plen, "cut::p20]p50]p70]p90]"))
#>
#> [1.0; 1.5] [1.6; 4.3] [4.4; 5.0] [5.1; 5.8] [5.9; 6.9]
#> 37 38 33 29 13
# Mixing all
table(bin(plen, "cut::2[q2]p90]"))
#>
#> [1.0; 1.9] [3.0; 4.3] [4.4; 5.8] [5.9; 6.9]
#> 50 25 62 13
# Adding custom names
table(bin(plen, c("cut::2[q2]p90]", "<2", "]2; Q2]", NA, ">90%")))
#> <2 ]2; Q2] [4.4; 5.8] >90%
#> 50 25 62 13
bin also accepts formulas, e.g. bin = list("<2" = ~ x < 2) (x must be the only variable).
bin accepts the use of .() for list().
you can add the location of the element using @d in the name. Useful to rearrange factors:
base = setNames(iris, c("y", "x1", "x2", "x3", "species"))
table(base$species)
#> setosa versicolor virginica
#> 50 50 50
table(bin(base$species, .("@3" = "seto", "@1 VIRGIN" = "virg")))
#> VIRGIN versicolor setosa
#> 50 50 50
the tex output is now "nicely" formatted.
argument extralines replaces the argument extraline to increase coherence. Hence function extraline_register becomes extralines_register (the change is done without deprecation since I guess this function must be only rarely used).
arguments extralines and headers accept .() for list().
base = setNames(iris, c("y", "x1", "x2", "x3", "species"))
check_conv_feols: checks the convergence of the fixed-effects in feols models by looking at the first-order conditions.Although a bit unrelated to the purpose of this package, these functions are so extensively used in the author's research that he decided to leverage his author privileges to include them in fixest to make them easier to share with co-authors.
osize: simple function returning a formatted object size.
n_unik: simple but flexible function returning the number of unique elements from variables in one or several data sets. Useful for checking keys.
sample_df: simple function to extract random lines from a data.frame.
when computing Newey-West standard-errors for time series, the bandwidth is now selected thanks to the bwNeweyWest function from the sandwich package. This function implements the method described in Newey and West 1994.
add type = "se_long" to summary.fixest_multi which yields all coefficients and SEs for all estimations in a "long" format.
only in fixest estimations, using a "naked" dot square bracket variable in the left-hand-side includes them as multiple left hand sides. Regular expressions can also be used in the LHS.
base = setNames(iris, c("y", "x1", "x2", "x3", "species"))
y = c("y", "x1")
feols(.[y] ~ x2, base)
#> Standard-errors: IID
#> Dep. var.: y
#> Estimate Std. Error t value Pr(>|t|)
#> (Intercept) 4.306603 0.078389 54.9389 < 2.2e-16 ***
#> x2 0.408922 0.018891 21.6460 < 2.2e-16 ***
#> ---
#> Dep. var.: x1
#> Estimate Std. Error t value Pr(>|t|)
#> (Intercept) 3.454874 0.076095 45.40188 < 2.2e-16 ***
#> x2 -0.105785 0.018339 -5.76845 4.5133e-08 ***
etable(feols(..("x") ~ y + i(species), base))
#> model 1 model 2 model 3
#> Dependent Var.: x1 x2 x3
#>
#> (Intercept) 1.677*** (0.2354) -1.702*** (0.2301) -0.4794** (0.1557)
#> y 0.3499*** (0.0463) 0.6321*** (0.0453) 0.1449*** (0.0306)
#> species = versicolor -0.9834*** (0.0721) 2.210*** (0.0705) 0.9452*** (0.0477)
#> species = virginica -1.008*** (0.0933) 3.090*** (0.0912) 1.551*** (0.0617)
#> ____________________ ___________________ __________________ __________________
#> S.E. type IID IID IID
#> Observations 150 150 150
#> R2 0.56925 0.97489 0.93833
#> Adj. R2 0.56040 0.97438 0.93706
improve error messages when subset does not select any element.
in xpd and fixest estimations, variables can be "grepped" from the data set with regex("regex").
add inheritance of the default style in iplot when the style is set globally with setFixest_coefplot.
improve error messages in general by prompting additional error calls (when appropriate).
the dictionaries now ignore white spaces in coefficient names (thanks to Caleb Kwon).
the package startup messages have been improved (they should pop up less often).
to comply with CRAN policies, the startup message doesn't write on the .Renviron file any more.
Fix bug occurring in IV models with multiple instruments and with multithreading on. That bug could lead to the wrong imputation of the IV residuals, hence affecting the standard-errors (although the order of magnitude of the variation should be minor). Thanks to @whitfillp, #182.
Fix minor, rare, bug occurring in feglm when the model was badly specified and VAR(Y) >>>> VAR(X) and there were only one variable.
model.matrix did not work with type = "fixef" (thanks to @kylebutts, #172).
In nonlinear estimations:fixef.rm = "none" or fixef.rm = "singleton" did not work as expected (thanks to @kre32, #171).
Fix bug that could occur when observations had to be removed on several fixed-effects dimensions (had no impact on the estimates though).
Fix bug in etable when file is provided and tex = FALSE (thanks to @roussanoff, #169).
Fix bug when: i) a fixest_panel is used as a data set in an estimation, ii) NA values are to be removed and iii) fixed-effects are used. Thanks to Nicola Cortinovis for the report!
Fix bug in to_integer when converting multiple vectors and sorting is required, without items.
Fix bug in feols.fit when the matrix of regressors was only partially named (reported by @leucothea, #176).
Fix bug in the value of the fixed-effects coefficients in IV estimations (thanks to @tappek, #190).
Fix bug in coefplot when lean = TRUE in the estimation (reported by @adamaltmejd, #195).
Fix bug in iplot when IVs contained interactions.
Fix bug in iplot preventing some variables to be removed (reported by @roussanoff, #164).
Fix 0 right-padding of numbers displayed in estimation results that could be confusing (reported by @hjuerges, #197).
New argument vcov:
greatly simplifies and extends how to specify the SEs
completely replaces the arguments se and cluster (they still work)
accepts functions
see the documentation in the vignette
New built-in VCOVs:
Newey-West (1987) for serially correlated errors
Driscoll-Kraay (1998) for cross-sectionally and serially correlated errors
Conley (1999) for spatially correlated errors
you can summon variables from the environment directly in the formula using the new dot square bracket (DSB) operator. The DSB operator can be used to create many variables at once, and can also be using within regular expressions. One example:
base = setNames(iris, c("y", "x1", "x2", "x3", "species"))
i = 2:3
z = "i(species)"
feols(y ~ x.[i] + .[z], base)
#> OLS estimation, Dep. Var.: y
#> Observations: 150
#> Standard-errors: IID
#> Estimate Std. Error t value Pr(>|t|)
#> (Intercept) 3.682982 0.107403 34.291343 < 2.2e-16 ***
#> x2 0.905946 0.074311 12.191282 < 2.2e-16 ***
#> x3 -0.005995 0.156260 -0.038368 9.6945e-01
#> species::versicolor -1.598362 0.205706 -7.770113 1.3154e-12 ***
#> species::virginica -2.112647 0.304024 -6.948940 1.1550e-10 ***
#> ---
#> Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#> RMSE: 0.333482 Adj. R2: 0.832221
in i and sunab you can now bin the variables on the fly with the new argument bin. The new function bin is also available at the user-level.
Function dof has been renamed into ssc (stands for small sample correction) to improve clarity. Retro compatibility is partially ensured but the function dof will be removed at some point.
Functions setFixest_dof and setFixest_se have been renamed into setFixest_ssc and setFixest_vcov. No retro compatibility ensured.
Removal of the var::factor operator to interact a continuous variable with a variable treated as a factor.
feglm now accepts partially matched character shortcuts for families: "poisson", "logit", "probit" are now valid family arguments.
predict.fixest accepts the new argument fixef which, if TRUE, returns a data.frame of the fixed-effects coefficients for each observation, with the same number of columns as the number of fixed-effects (feature requests #144 and #175 by @pp2382 and @cseveren).
offsets present in the formula are now accepted.
VCOV aliases (both Grant McDermott's suggestions: thanks Grant!):
the default standard-errors is now "iid" (former keywords still work)
the keyword hc1 can be used to summon heteroskedasticity-robust SEs
Argument sliding: the argument vcov can be called implicitly when data is set up globally:
base = setNames(iris = c("y", "x1", "x2", "x3", "species"))
# Setting up the data
setFixest_estimation(data = base)
# Now vcov can be used without using vcov = stuff:
feols(y ~ x1 + x2, ~species)
# => same as feols(y ~ x1 + x2, vcov = ~species)
mtcars |> feols(cyl ~ mpg)
# => same as feols(cyl ~ mpg, mtcars)
the user can now specify custom degrees of freedom to compute the t-tests in ssc() (feature request by Kyle F. Butts).
the predict method gains the new argument se.fit and interval which computes the SEs/CI of the predicted variable. This only works for OLS models without fixed-effects. Feature request by Gábor Békés, #193.
iplot gains the argument i.select to navigate through the different variables created with i() (provided there are more than one of course).
new argument interaction.order to control the order in which interacted variables are displayed (feature request by @inkrement, #120).
new argument i.equal to control how the values taken by factor variables created with i() are displayed.
new meta.XX family of arguments when exporting to Latex. They include various type of information as comments before the table (suggestion of adding the time by Apoorva Lal, #184). So far the new arguments are: meta.time, meta.author, meta.sys, meta.call, meta.comment. The argument meta is a shortcut to all these.
default values can be saved at the project level using the argument save = TRUE in the function setFixest_etable. This means that default values will be automatically set without having to call setFixest_etable at the startup of the R session. For example, if you want to permanently add the creation time in your Latex exports, just use setFixest_etable(meta.time = TRUE, save = TRUE), and you won't need to bother about it anymore in any future session in your current project.
some changes in extraline, now: i) it accepts raw vectors, ii) it accepts lines without title, iii) the elements are recycled across models, iv) it accepts elements of the form list("item1" = #item1, "item2" = #item2, etc), and v) the elements are Latex-escaped.
in style.tex: all Latex-escaping is removed.
the argument subtitle has been renamed into headers (retro-compatibility is ensured).
on the new argument headers:
headers = list("Gender" = list("M" = 3, "F" = 4)) will create a line with 3 times "M" and 4 times "F".":_:" in the row name will add a rule for each column group (in the previous example ":_:Gender" would do). Suggestion by @nhirschey, #173.headers = list("_Gender" = list("M" = 3, "F" = 4)) will place the header line at the very bottom of the headers.":tex:" in the row title.argument sdBelow has been renamed into se.below (retro-compatibility is ensured).
new argument se.row to control whether the row displaying the standard-errors should be displayed (clarification requested by @waynelapierre, #127).
dict now directly modifies the entries in the global dictionary instead of creating a brand new one. For example if you have setFixest_dict(c(cyl="Cylinder")) and then use etable with dict=c(mpg="miles per gallon"), you end up with both the names cyl and mpg to be modified. To disable this behavior, you can add "reset" as the first element, like in dict=c("reset", mpg="miles per gallon").
Major bug, leading R to crash, occurring when the same variable was used with several different slopes (thanks to @Oravishayrizi, #119).
Major bug, leading R to crash, occurring when 3+ fixed-effects are to be combined.
Major bug, leading R to crash, occurring when multiple LHS are estimated with the option fixef.rm = "singleton" (thanks to Ole Rogeberg).
Major bug, leading R to crash, occurring when many fixed-effects have to be removed because of only 0/1 outcomes (thanks to @mangelett #146 and @ChristianDueben #157).
Fix bug occurring for undefined covariances with only one regressor (thanks to @joseph-richard-martinez, #118).
Fix bug in IV estimations regarding the Wald statistic of the first stage when lean = TRUE and the VCOV computation is done post estimation.
Fix bug in the Wald test in IV estimations when variables are removed because of collinearity (thanks to @pei-huang, #117).
Fix bug regarding multiple estimations when the multiple fixed-effects contained variables with varying slopes.
Fix various display bugs in fitstat.
Fix bug in etable: using split sample estimations prevented the argument title to render correctly.
Fix incorrect information message when observations are removed because of infinite values (in some circumstances the removal was wrongly attributed to NAness).
Fix bug in etable when checking the argument coefstat (thanks to @waynelapierre, #121).
Fix bug in feols when IV estimations contained fixed-effects and lean = TRUE (thanks to @adamaltmejd, #123).
Fix bug in IV estimations when an endogenous regressor was removed because of collinearity.
Fix bug estimation without intercept not working when lags are present in the formula (thanks to @nreigl, #126).
Fix various bugs when using subset in the estimations (reported by @noahmbuckley and @Oravishayrizi, #129 and #131).
Fix error message when data cannot be fetched (reported by @Oravishayrizi, #134).
Fix bug getting the "G" statistic in fitstat.
Fix bug in predict when a poly() term was used and the formula was long (reported by @XiangLiu-github, #135).
fix bug for extracting sub statistics of "ivwald" and "ivf" in fitstat.
fix bug when i() was used without intercept.
Fix display bug in etable when Tex output is requested and interactions composed of identical variables with different interacted orders are present (reported by @Oravishayrizi, #148).
Fix bug in etabe when fixef.group is used and fixed-effects are renamed (reported by @jamesfeigenbaum).
Fix bug when fplit is used with subset.
Fix bug when using cluster with subset and NA values are removed (reported by @adamaltmejd, #154).
Fix bug argument lean not working on summary when applied to an existing summary and only the argument lean was present (reported by @adamaltmejd).
Fix bug when using multiple LHS with lags in the formula (reported by @Nicolas Reigl, #158).
Fix bug regarding the intercept-only likelihood when weights are provided (only with Poisson and logit models), reported by @fostermeijer, #155.
the function i(), used to create factors or interactions has been tidied up, leading to breaking changes.
the first two arguments have been swapped! such that now the first argument will always be treated as a factor.
the new syntax is i(factor_var, var, ref, keep, ref2, keep2) where var can be either continuous or factor-like (the argument f2, for interaction with factors, has been removed).
Fix rare bug when the number of parameters is greater than the number of observations and the GLM family has a dispersion parameter.
glm, the new default family for feglm is gaussian (previously it was Poisson, if you were using it, please now use the function fepois instead).the function coefplot has been split in two:
coefplot: always plots all the coefficients.
iplot: plots only interactions or factors created with the function i().
the function iplot hence replaces coefplot's former argument only.inter which controlled whether or not to focus on interactions.
group and extraline: enhanced and simplified control of the placement of the new rows. Now only two special characters at the beginning of the row name decide its location.
new argument fixest.group. If TRUE, then fixed-effects appearing always jointly across models will be grouped in a single row. The user can alternatively specify a list to declare which fixed-effect to group and customize the row name.
sdBelow now works when tex = FALSE (request by Sasha Indarte).
extraline can now be equal to a formula containing extraline macros or valid fitstat types.
When a list, extraline can contain functions (returning a scalar) that will be applied to each model.
When a list, extraline can contain formulas containing extraline macros or valid fitstat types.
You can register extraline macros with the new function extraline_register.
When tex = TRUE, n-way clustering now always leads to the name of clustered SEs (n-way is not shown any more).
Add the argument coef.just that controls the justification of the coefficients and standard-errors. Only works when tex = FALSE (i.e. a data.frame is requested).
new function sunab that simplifies the implementation of the SA method.
just type sunab(cohort, period) in a fixest estimation and it works!
Common methods have been extended to fixest_multi objects.
coef.fixest_multi: re-arranges the coefficients of multiple estimations into a matrix.
resid.fixest_multi: re-arranges the residuals of multiple estimations into a matrix.
kpr: Kleibergen-Paap rank test for IV estimations.
cd: Cragg-Donald F statistic for IV estimations.
my: gives the mean of the dependent variable.
degrees_freedom: to access the DoFs of the models (sometimes that can be intricate).
feols.fit: fit method for feols.
obs: to obtain the observations used in the estimation.
All fixest estimation now accept scalars from the global environment (variables are still not allowed!).
Better handling of the DoFs in fitstat (in particular when the VCOV is clustered).
model.matrix:
the endogenous and exogenous regressors, and the instruments from IV estimations can now be easily extracted.
new arguments as.matrix and as.df to coerce the result to a particular format.
.fit methods (feols.fit and feglm.fit) now handle multiple dependent variables.
to_integer now sorts appropriately any kind of vectors (not just numeric/character/factors).
substantial speed improvement when combining several vectors with many cases (> millions).
The number of threads to use can now be set permanently at the project level with the new argument save in the function setFixest_nthreads.
model.matrix when type = "lhs" the return object is now a vector (previously it was a data.frame).Improve error messages.
hatvalues.fixest: now returns an error instead of a message when fixed-effects are present (it makes the interplay with sandwich 'nicer').Fix bug depvar = FALSE not working when tex output was requested (thanks to @apoorvalal and @pbaylis, #104).
Fix bug in naming when i() led to only one variable being retained (thanks to @
colejharvey, #106).
Fix bug display when only degrees of freedom are selected in fitstat.
Fix bug when lean = TRUE in IV estimations with fixed-effects (a large object was still present, thanks to @zozotintin).
Fix bug display of etable in Rmarkdown (thanks to @kdzhang, #93, and @nikolassch, #112)
Improve error messages in fitstat when selecting statistics components.
Fix bug in predict when poly() was used in the estimation (thanks to @tholdaway, #109).
Fix bug in predict: an error message would not pop when combined fixed-effects are used with combine.quick = TRUE (thanks to @benzipperer, #115).
Fix bug to properly account for the nestedness of combined fixed-effects when clustered standard-errors are requested (thanks to @Oravishayrizi , #116).
Fig major bug in model.matrix that could make it very slow. Led the function aggregate to be very slow (thanks to Benny Goldman).
Fix bug that prevented aggregate to effectively use weights (thanks to Benny Goldman).
model.matrix gains the new argument subset which allows the creation of the design matrix for a subset of variables only.
drop.section now works for etable when tex = FALSE.
The argument panel.id used in all estimations can be set globally with the function setFixest_estimation.
i: Factor variables with only the values of 0 and 1 are treated as numeric.
fitstat: The statistic G is now equal to the degrees of freedom used in the t-test of coefficients testing.
esttable and esttex are not deprecated any more: they are now pure aliases of etable.
aggregate: for weighted regressions, use_weights controls whether to use the weights to perform the aggregation.
Remove test that leads to a (uber odd) bug on fedora devel.
Fix bug in IV estimation when using factors as instrumented variables (thanks to @adamaltmejd, #99).
Fix bug when using at least two fixed-effects and varying slopes with singletons (thanks to @adamtheising, #89).
xpd, macros are parsed even after creating the formula with the lhs and rhs arguments.Fix bug in IV estimations when lean = TRUE (thanks to @reifjulian, #88).
Fix various bugs related to the use of summary when lean = TRUE in the estimation.
Fix bug preventing se = "cluster" to be used in etable (thanks to Caleb Kwon).
Fix bug etable not escaping variable names properly when sdBelow = FALSE (thanks to Jeppe Viero).
Fix bug in IV estimation with lean = TRUE.
Fix bug preventing the return of demeaned variables in IV estimations (thanks to @amarbler, #94).
i() now automatically converts its first argument to numeric if it was of type logical. The user can still pass logicals to the argument f2 if the expected behavior is really to treat it as a logical.
Improve fitstat help and error messages.
Bug in etable when the default value of fitstat was set with setFixest_etable.
Bug in model.matrix when the model contained fixed-effects and the RHS was requested: the intercept was wrongfully added.
Fix rare bug when i() was called within a very specific set of functions.
Fix bug in R old release due to anyNA.data.frame.
Fix bug regarding panel data sets when variables were created in a data.table within functions (thanks to @tcovert, #76).
Add extra elements to be removed when lean = TRUE to keep the object as small as possible (reported by @zozotintin, #81).
Fix bug in fixed-effects estimations with multiple LHS and different number of observations per estimation that prevented to get the default behavior for standard-errors to work.
Fix occasional bug when using split with fixed-effects.
xpd now appropriately returns a two sided formula when a one sided formula is fed in and the argument lhs is provided.
Fix bug in coefplot preventing the proper scaling of the x-axis for interactions when multiple models are displayed.
Fix occasional bug in the ordering of sub-selections of multiple estimations.
For staggered difference-in-difference analyzes: the method of Sun and Abraham (forthcoming, Journal of Econometrics) has been implemented.
After having used i() to interact cohort dummies with time to treatment dummies, use the function aggregate to recover the yearly treatment effects.
So far the way to do it, although easy, is a bit arcane but the next versions of the software will include a user-friendly way.
For details, check out the help page of the function aggregate or the staggered DiD section in the vignette fixest walkthrough.
Function i() now has the new arguments f2, drop2 and keep2 which allows the interaction of two factors (useful for staggered DiD estimations).
Argument dof, used to compute the standard-errors, can now be used at estimation time.
In etable, the argument digits can now accepts a character value specifying the way the decimals should be displayed. For example if digits = "r2" this means that all numbers will be rounded at two decimals and these two decimals will always be displayed. The default behavior is to display significant digits. Follows feature request #82 by @lyifa.
etable also gains the argument digits.stats which monitors how the fit statistics decimals should be displayed.
Argument split now accepts variable names.
More coherence regarding the use of summary applied to models for which the SEs were computed at estimation time. Now there is a memory of how the SEs were computed, so that, for example, if only the argument dof is passed to summary, then the SEs will be clustered in the same way as estimation time and only dof will change.
Now an error is raised when i() is used in the fixed-effects part of the formula. The appropriate way is indicated (related to #77 by
@rrichmond).
Improved default setting of standard-errors.
Improved error messages.
In multiple estimations, models returning full NA coefficients are not returned (instead of raising an error).
Major bug when predict was used in the presence of fixed-effects (thanks to @jurojas5, #54). Introduced in version 0.7.
When using variable names to cluster the standard-errors inside functions, summary may not fetch the data in the right frame (thanks to @chenwang, #52). Now a completely new internal mechanic is in place.
When using variables with varying slopes and the number of iterations is greater than 300, a bug occurred in the function checking the convergence was right (thanks to @kendonB, #53).
Fix bug in the demeaning algorithm when two variables with varying slopes were identical.
Fix bug in femlm/feNmlm when factor variables are removed due to the removal of some observations.
In summary, fix bug when the argument cluster was equal to a formula with expressions and not a variable name (thanks to @edrubin, #55).
Fix bug when integers are present in the RHS (thanks to @zozotintin, #56).
Fix bug when nb_FE >= 2 and the data was large (thanks to @zozotintin, #56).
Fix bug display of how the standard-errors were clustered in etable.
Fix bug occurring when lags were used in combination with combined fixed-effects (i.e. fe1 ^ fe2) (thanks to @SuperMayo, #59).
Fix bug coefplot when representing multiple estimations and coefficient names are numbers.
base = iris
names(base) = c("y", "x1", "x_endo", "x_inst", "species")
base$endo_bis = 0.5 * base$y + 0.3 * base$x_inst + rnorm(150)
base$inst_bis = 0.2 * base$x_endo + 0.3 * base$endo_bis + rnorm(150)
# The endo/instrument is defined in a formula past a pipe
res_iv1 = feols(y ~ x1 | x_endo ~ x_inst, base)
# Same with the species fixed-effect
res_iv2 = feols(y ~ x1 | species | x_endo ~ x_inst, base)
# To add multiple endogenous regressors: embed them in c()
res_iv3 = feols(y ~ x1 | c(x_endo, x_endo_bis) ~ x_inst + x_inst_bis, base)
The fitstat function has been significantly enhanced.
Now the following types are supported:
Likelihood ratios
F-tests
Wald tests
IV related tests (F/Wald/Sargan)
common stats like the R2s, the RMSE, Log-likelihood, etc
You can register your own fit statistics. These can then be seamlessly summoned in etable via the argument fitstat.
The print.fixest function now supports the fitstat argument. This means that you can display your own desired fit statistics when printing fixest objects. This is especially useful in combination with the setFixest_print function that allows to define the default fit statistics to display once and for all. See the example in the "Instrumental variables" section of the Walkthrough vignette.
The new function wald computes basic Wald tests.
New arguments split and fsplit: you can now perform split sample estimations (fsplit adds the full sample).
Estimations for multiple left-hand-sides can be done at once by wrapping the variables in c().
In the right-hand-side and the fixed-effects parts of the formula, stepwise estimations can be performed with the new stepwise functions (sw, sw0, csw and csw0).
The object returned is of class fixest_multi. You can easily navigate through the results with its subset methods.
aq = airquality[airquality$Month %in% 5:6, ]
est_split = feols(c(Ozone, Solar.R) ~ sw(poly(Wind, 2), poly(Temp, 2)),
aq, split = ~ Month)
# By default: sample is the root
etable(est_split)
# Let's reorder, by considering lhs the root
etable(est_split[lhs = TRUE])
# Selecting only one LHS and RHS
etable(est_split[lhs = "Ozone", rhs = 1])
# Taking the first root (here sample = 5)
etable(est_split[I = 1])
# The first and last estimations
etable(est_split[i = c(1, .N)])
..("regex"):
data(longley)
# All variables containing "GNP" or "ployed" in their names are fetched
feols(Armed.Forces ~ Population + ..("GNP|ployed"), longley)
New style.tex and style.df arguments that define the look of either Latex tables or the output data.frames.
it can be set with the new functions style.tex and style.df that contain their own documentation.
some etable arguments have been ported to the style functions (yesNo, tablefoot).
New postprocess.tex and postprocess.df arguments which allow the automatic postprocessing of the outputs. See the dedicated vignette on exporting tables for an illustration.
new tabular arguments which allows to create tabular* tables (suggestion by @fostermeijer, #51).
polynomials and powers are automatically renamed to facilitate comparison across models. You can set their style with the argument poly_dict.
the labeling of models is enhanced when rep.fixest is used with different standard-errors (the model names are now "model INDEX.SUB-INDEX").
the argument subtitles has been improved, and now automatically displays the samples when split sample estimations are performed.
In all estimations:
subset: regular subset (long overdue).
split, fsplit: to perform split sample estimations.
se, cluster: to cluster the standard-errors during the call.
lean: if TRUE, then summary is applied and any large object is removed from the result. To save memory => but many methods won't work afterwards.
fixef.rm: argument that accepts none, perfect, singleton, both. Controls the removal of fixed-effects from the observation.
auto parsing of powers. Now you don't need to use I() to have powers of variables in the RHS, it is automatically done for you (i.e. x^3 becomes I(x^3)):
base = iris
names(base) = c("y", "x1", "x2", "x3", "species")
# The multiple estimation below works just fine
feols(y ~ csw(x, x^2, x^3), base)
Estimation options can be set globally with setFixest_estimation().
The demean function has been enhanced (with the contribution of Sebastian Krantz).
Internal demeaning algorithm: some copies of the data are avoided when using feglm.
Internal algorithm of to_integer (used in all estimations): one copy of the input data is now avoided.
All estimations: smarter handling of the intercept, thus avoiding the reconstruction of the design matrix.
Fix bug int overflow in estimations with only one variable.
Fix bug in tests occurring in R old release.
Fix bug in examples occurring in R old release.
Function i() now behaves as factor(), setting automatically a reference when appropriate.
Internal algorithm of i() is much faster.
In etable, the user can now provide a type of clustering for each model.
New method rep.fixest to replicate fixest objects, mostly useful in etable when several SEs for the same models are to be reported.
Automatic fix when the variance is not positive definite.
Major bug when fixed-effects were combined with ^ and they contained NAs (thanks to @poliquin, #35).
Bug when using lead/lags in estimations. The bug was due to a bug in a dependency (dreamerr) and was fixed. Now fixest requires dreamerr version >= 1.2.1. Bug spotted by @seunghoon001 (#44).
Major bug when n_obs x n_vars > 2B or n_obs x n_fixed-effects > 2B. In such cases estimations could just not be done, even leading R to crash when using nthreads > 1. The algorithm was fixed to allow datasets with up to 2B observations to be estimated in all circumstances. Bug reported, and many help for checking provided, by Howard Zihao Zhang.
coefplot: Problem regarding interactions when observations, and hence coefficients, were removed from the estimation. Now the coefficients are removed from the plot. Bug reported by @phisherblack, #45.
coefplot: Corrected various bugs when asked for the plotting of several estimations.
Brand new internal algorithm which now uses closed form solutions when dealing with variables with varying slopes. This means that when variables with varying slopes are present, the algorithm is incomparably faster and more accurate.
Two deep copies of some data are now avoided in the demeaning function. This improves the performance in terms of memory footprint, and also makes the algorithm faster.
New default values for standard-errors (only concerns multiway clustering). They become similar to reghdfe to increase cross-software comparability. Computing the standard-errors the old way is still possible using the argument dof. See the dedicated vignette: On standard errors.
Name change in summary/vcov/etable: To get heteroskedasticity-robust standard-errors, se = "hetero" now replaces se = "white" to enhance clarity. Note that se = "white" still works.
fitsat that computes various fit statistics. It is integrated with etable and can be invoked with the argument fitstat. So far only two fit statistics are included, but more will come.You can now use i(var) to treat the variable var as a factor. You can select which values to drop/keep with the respective arguments.
i(var) leads to a special treatment of these variables in the functions coefplot and etable.New argument placement to define the position of the float in Latex (suggestion by Caleb Kwon).
New argument drop.section, with which you can drop a) the fixed-effects, b) the variables with varying slopes, or c) the statistics, sections (suggestion by Caleb Kwon).
Fix glitch in help pages regarding the use of the '%' (percentage) character in regular expressions.
Two new arguments .vcov and .vcov_args to compute the standard-errors with custom functions.
The number of observations (n) is now treated as a regular statistic and can be placed where one wants.
The statistics can now have custom aliases using the argument dict.
The overdispersion becomes a regular fit statistic that can be included (or not) using fitstat.
The dictionnary now applies to the factors of interactions, and the values of factors.
Argument nthreads:
The new default of argument nthreads is 50% of all available threads.
Accepts new values: a) 0 means all available threads, b) a number strictly between 0 and 1 will represent the fraction of all threads to use.
When setting formula macros:
xpd and setFixest_fml now accept character vectors and numeric scalars on top of formulas.demean:
coefplot:
The argument group now accepts a special character "^^", when used, it cleans the beginning of the coefficient name. Very useful for, e.g., factors although factors created with i() need not that.
When horiz = TRUE, the order of the coefficients is not reversed any more.
Improved display of numbers in print method.
Added variables names to X_demeaned from feols.
Lagging functions:
Now time.step = NULL by default, which means that the choice of how to lag is automatically set. This means that the default behavior for time variables equal to Dates or character values should be appropriate.
New operator d which is the difference operator.
In all estimations:
mem.clean: internally, intermediary objects are removed as much as possible and gc() is called before each memory intensive C++ section. Only useful when you're at the edge of reaching the memory limit.collin.min_norm, this value informs on the possible presence of collinearity in the system of variables.only.env and env:
only.env, allows to recover only the environment used to perform the estimation (i.e. all the preprocessing done before the estimation).env, accepts a fixest environment created by only.env, and performs the estimation using this environment--all other arguments are ignored.env, we cut all preprocessing).In non-linear estimations:
NL.start now accepts numeric scalars, initializing all coefficients to the same value (avoids the use of the other argument NL.start.init).summary.fixest:
.vcov now accepts functions that compute the vcov. This ensures convenient compatibility with the sandwich package (compatibility is still not full though: bootstraped SEs don't work yet).update.fixest:
evaluate to ensure consistency with the update method from stats.feols & feglm:
na_inf.rm has been removed. It was present for historical reasons, and is removed to increase code clarity.In vcov, the degree-of-freedom in the small sample correction correction was fixed to "nested" and couldn't be modified, now corrected. Further, "nested" was not properly accounted for, now corrected.
In etable, fitsat = FALSE or fitsat = NA led to a bug.
r2: bug when the estimation contained only fixed effects (thanks to Luis Fonseca, #27).
Now the BIC of feglm is similar to the one of glm.
Bug in the log-likelihood in the presence of weights, now corrected.
Bug in coefplot when some interacted variables were removed because of collinearity. Now corrected.
On standard-errors: how are the SEs computed in fixest and how to replicate the SEs from other software.
Exporting estimation tables: how to use efficiently etable, in particular how to customize the tables.
New arguments: group, extraline, notes, tablefoot.
group allows to eliminate variables (like drop) and adds an extra line with TRUE/FALSE if the model contained those variables.
extraline allows to add extra lines with any content.
notes allows to add notes after the table (suggestion by @bgchamps, #25).
tablefoot controls whether the table footer, containing the type of standard-errors and the significance codes, should be displayed.
Renaming: yesNoFixef => yesNo.
Most default values can be set globally with the new function setFixest_etable.
dof, used to adjust the small sample corrections, is now much more complete and allows to replicate a large set of estimation results from alternative software.You can now provide custom VCOVs to summary by using the argument .vcov.
A warning is now prompted when the maximum number of iterations of the algorithm is reached (suggestion by @clukewatson , #24]).
The types of standard-errors can now be set globally with the function setFixest_se (suggestion by @dlindzee, #28)
New feols argument demeaned. If TRUE, then the centered variables are returned (y_demeaned and X_demeaned). (Suggestion by Linus Holtermann.)
interact gains two new arguments: drop and keep (suggestion by @SuperMayo, #23).
hatvalues has been implemented for feols and feglm estimations.
the estfun from sandwich has been implemented.
residuals.fixest.style which allows to set many elements of the output table.signifCode can be equal to "letters" to display letters instead of stars.setFixest_nthreads now respects the OMP_THREAD_LIMIT environment variable.var::fe syntax with confirm = TRUE and no reference.etable when the standard-errors where NA.feglm for non-poisson, non-binomial families, are now correct (minor differences).fixef did not work when the slope was an integer, now corrected (thanks to @clerousset, #20).setFixest_fml(..ctrl = ~ var1 + var2). Here the macro variable ..ctrl has been set to the value "var1 + var2".fixest estimation: e.g. data(airquality) ; setFixest_fml(..ctrl = ~ Temp + Day) ; feols(Ozone ~ Wind + ..ctrl, airquality).xpd, which expands formulas. E.g. lm(xpd(Ozone ~ Wind + ..ctrl), airquality).to_integer: user-level version of the internal algorithm transforming any kind of vector (or combination of vectors) into an integer ranging from 1 to the number of unique elements of the vector. Very fast.demean: user-level version of the demeaning algorithm used in feols.New internal algorithm to estimate OLS (applies to both feols and feglm):
It is numerically more stable.
Incomparably faster when factors are to be estimated (and not explicitly used as fixed-effects).
Collinear variables are removed on the fly.
var::fe(ref) now accept multiple references (i.e. ref can be a vector).etable, the variable names of non-Latex output can now be changed.n when applying summary to choose the number of coefficients to display.confirm has been removed from the function interact.r2 allows more flexibility in the keywords it accepts.dof gains a new argument adj which allows to make different types of common small sample corrections. Its other arguments have been renamed for clarity (fixef => fixef.K, exact => fixef.exact, cluster => cluster.adj).feols and non-poisson, non-binomial models in feglm. For all other models, z-statistics are used. This complies with the default's R-stats behavior.residuals method has been substantially improved, now allowing different types.collinearity help pages: an example could lead to an error (due to random data generation). It has been removed.collinearity, corrected the problem of display of the intercept in some situations.cex and lwd in coefplot have been changed to 1 and 1 (instead of par("cex") and par("lwd")). Otherwise this led to the creation of Rplots.pdf in the working directory (thanks to Kurt Hornik).fixef_sizes.simplify, which provides the sizes of the fixed-effects in parentheses when there is no ambiguity.signifCode = NA.float which decides whether to embed the table into a table environment. By default it is set to TRUE if a title or label is present.keep to select the variables to keep in the table.coefstat defining what should be shown below the coefficients (standard-errors, t-stats or confidence intervals). Suggestion by @d712, #16.horiz. The coefficients can now be displayed horizontally instead of vertically.lab.fit: "simple", the classic axis, "multi", the labels appear across multiple lines to avoid collision, and "tilted" for tilted labels.style allows you to set styles with the function setFixest_coefplot, you can then summon the style in coefplot with this argument.coefplot.group and group.par).terms.fixest giving the terms of the estimation.donttest sections were removed from help pages.etable. In the process, some of their arguments were "lost", this is now corrected.coefplot now accepts lists of estimations.You can now add lags and leads in any fixest estimations. You only need to provide the panel identifiers with the new argument panel.id, then you're free to use the new functions l() for lags and f() for leads.
You can also set up a panel data set using the function panel which allows you to use the lagging functions without having to provide the argument panel.id, and which dispose of more options for setting the panel.
You can now add interactions in formulas with a new syntax: var::fe(ref)
The command var::fe(ref) interacts the variable var with each value of fe and sets ref as a reference. Note that if you don't use the argument ref, the command var::fe is identical to var:factor(fe).
Using var::fe(ref) to write interactions opens up a special treatment of such variables in the exporting function etable and in the coefficient plotting function coefplot.
You can plot coefficients and their associated confidence intervals with the function coefplot.
coefplot dispose of many options, whose default values can be set with the function setFixest_coefplot.
As for the function etable, you can easily rename/drop/order the coefficients.
coefplot detects when interactions have been used and offers a special display for it.
[etable] Estimations table: new function to export the results of multiple estimations. Replaces the two functions esttex and esttable (the two functions still exist but they will be deprecated in the future).
[Lagging] New functions related to lagging: l, f, panel, unpanel and [.fixest_panel.
[Utilities] A set of small utility functions has been added. They allow to extract part a coefficient table or parts of it (like the t-statistics of the standard-error) from an estimation. These functions are coeftable, ctable (an alias to coeftable), se, tstat and pvalue.
[coefplot] The functions coefplot and setFixest_coefplot.
[dof] New function to set the type of degree of freedom adjustment when computing the variance-covariance matrix. You can permanently set the type of DoF adjustment with the new function setFixest_dof().
dict=c(x1="Wind", x2="Rain"), with an estimation with the following variables 'x1', 'x2', 'x1:x2' will lead to the following aliases in Latex 'Wind', 'Rain' and 'Wind $times$ Rain'.yesNoFixef can be of length one, defaulting the second element to the empty string.did_estimate_yearly_effects.feglm.-[did_means] New function did_means to conveniently compare means of groups of observations (both treat/control and pre/post). Contains tools to easily export in Latex.
sym macro in Latex is dropped.This package is an effort to create a family of fast and user-friendly functions to perform estimations with multiple fixed-effects (F.E.).
Estimations with fixed-effects (or call it factor variables) is a staple in social science. Hence having a package gathering many methods with fast execution time is of prime importance. At the time of this version, this is the fastest existing method to perform F.E. estimations (often by orders of magnitude, compared to the most efficient alternative methods [both in R and Stata]). The underlying method to obtain the F.E. is based on Berge 2018, and the workhorse of the code is in c++ parallelized via OpenMP (btw thanks Rcpp for simplifying coders' life!).
This package is the follow up of the (now deprecated) package FENmlm which performed fixed-effects estimations but for only four likelihood families. Package fixest completely supersedes FENmlm by extending the method to regular OLS and all GLM families, and adding new utility functions. Further, the design of the functions has been completely overhauled and extended towards much more user-friendliness. Massive effort has been put into providing a set of informative error messages to the user for quick debugging of her workflow (e.g. one of the functions contains over 100 different errors messages).