emfx {etwfe} | R Documentation |
Post-estimation treatment effects for an ETWFE regressions.
Description
Post-estimation treatment effects for an ETWFE regressions.
Usage
emfx(
object,
type = c("simple", "group", "calendar", "event"),
by_xvar = "auto",
collapse = "auto",
post_only = TRUE,
...
)
Arguments
object |
An |
type |
Character. The desired type of post-estimation aggregation. |
by_xvar |
Logical. Should the results account for heterogeneous
treatment effects? Only relevant if the preceding |
collapse |
Logical. Collapse the data by (period by cohort) groups
before calculating marginal effects? This trades off a loss in estimate
accuracy (typically around the 1st or 2nd significant decimal point) for a
substantial improvement in estimation time for large datasets. The default
behaviour ("auto") is to automatically collapse if the original dataset
has more than 500,000 rows. Users can override by setting either FALSE or
TRUE. Note that collapsing by group is only valid if the preceding |
post_only |
Logical. Drop pre-treatment ATTs? Only evaluated if (a)
|
... |
Additional arguments passed to
|
Value
A slopes
object from the marginaleffects
package.
Performance tips
Under most situations, etwfe
should complete very quickly. For its part,
emfx
is quite performant too and should take a few seconds or less for
datasets under 100k rows. However, emfx
's computation time does tend to
scale non-linearly with the size of the original data, as well as the
number of interactions from the underlying etwfe
model. Without getting
too deep into the weeds, the numerical delta method used to recover the
ATEs of interest has to estimate two prediction models for each
coefficient in the model and then compute their standard errors. So, it's
a potentially expensive operation that can push the computation time for
large datasets (> 1m rows) up to several minutes or longer.
Fortunately, there are two complementary strategies that you can use to
speed things up. The first is to turn off the most expensive part of the
whole procedure—standard error calculation—by calling emfx(..., vcov = FALSE)
. Doing so should bring the estimation time back down to a few
seconds or less, even for datasets in excess of a million rows. While the
loss of standard errors might not be an acceptable trade-off for projects
where statistical inference is critical, the good news is this first
strategy can still be combined our second strategy. It turns out that
collapsing the data by groups prior to estimating the marginal effects can
yield substantial speed gains of its own. Users can do this by invoking
the emfx(..., collapse = TRUE)
argument. While the effect here is not as
dramatic as the first strategy, our second strategy does have the virtue
of retaining information about the standard errors. The trade-off this
time, however, is that collapsing our data does lead to a loss in accuracy
for our estimated parameters. On the other hand, testing suggests that
this loss in accuracy tends to be relatively minor, with results
equivalent up to the 1st or 2nd significant decimal place (or even
better).
Summarizing, here's a quick plan of attack for you to try if you are worried about the estimation time for large datasets and models:
Estimate
mod = etwfe(...)
as per usual.Run
emfx(mod, vcov = FALSE, ...)
.Run
emfx(mod, vcov = FALSE, collapse = TRUE, ...)
.Compare the point estimates from steps 1 and 2. If they are are similar enough to your satisfaction, get the approximate standard errors by running
emfx(mod, collapse = TRUE, ...)
.
Heterogeneous treatment effects
Specifying etwfe(..., xvar = <xvar>)
will generate interaction effects
for all levels of <xvar>
as part of the main regression model. The
reason that this is useful (as opposed to a regular, non-interacted
covariate in the formula RHS) is that it allows us to estimate
heterogeneous treatment effects as part of the larger ETWFE framework.
Specifically, we can recover heterogeneous treatment effects for each
level of <xvar>
by passing the resulting etwfe
model object on to
emfx()
.
For example, imagine that we have a categorical variable called "age" in
our dataset, with two distinct levels "adult" and "child". Running
emfx(etwfe(..., xvar = age))
will tell us how the efficacy of treatment
varies across adults and children. We can then also leverage the in-built
hypothesis testing infrastructure of marginaleffects
to test whether
the treatment effect is statistically different across these two age
groups; see Examples below. Note the same principles carry over to
categorical variables with multiple levels, or even continuous variables
(although continuous variables are not as well supported yet).
See Also
Examples
## Not run:
# We’ll use the mpdta dataset from the did package (which you’ll need to
# install separately).
# install.packages("did")
data("mpdta", package = "did")
#
# Basic example
#
# The basic ETWFE workflow involves two steps:
# 1) Estimate the main regression model with etwfe().
mod = etwfe(
fml = lemp ~ lpop, # outcome ~ controls (use 0 or 1 if none)
tvar = year, # time variable
gvar = first.treat, # group variable
data = mpdta, # dataset
vcov = ~countyreal # vcov adjustment (here: clustered by county)
)
# mod ## A fixest model object with fully saturated interaction effects.
# 2) Recover the treatment effects of interest with emfx().
emfx(mod, type = "event") # dynamic ATE a la an event study
# Etc. Other aggregation type options are "simple" (the default), "group"
# and "calendar"
#
# Heterogeneous treatment effects
#
# Example where we estimate heterogeneous treatment effects for counties
# within the 8 US Great Lake states (versus all other counties).
gls = c("IL" = 17, "IN" = 18, "MI" = 26, "MN" = 27,
"NY" = 36, "OH" = 39, "PA" = 42, "WI" = 55)
mpdta$gls = substr(mpdta$countyreal, 1, 2) %in% gls
hmod = etwfe(
lemp ~ lpop, tvar = year, gvar = first.treat, data = mpdta,
vcov = ~countyreal,
xvar = gls ## <= het. TEs by gls
)
# Heterogeneous ATEs (could also specify "event", etc.)
emfx(hmod)
# To test whether the ATEs across these two groups (non-GLS vs GLS) are
# statistically different, simply pass an appropriate "hypothesis" argument.
emfx(hmod, hypothesis = "b1 = b2")
#
# Nonlinear model (distribution / link) families
#
# Poisson example
mpdta$emp = exp(mpdta$lemp)
etwfe(
emp ~ lpop, tvar = year, gvar = first.treat, data = mpdta,
vcov = ~countyreal,
family = "poisson" ## <= family arg for nonlinear options
) |>
emfx("event")
## End(Not run)