Plot effects after estimating an autoregressive distributed lag (ARDL) model.
lrplot estimates and plots the period-specific and long-run effects of an autoregressive distributed lag (ARDL) model.
lrplot simulates the effects by taking draws from a multivariate normal distribution where the means of each variable
are the coefficients from the regression and the variance–covariance matrix is the variance–covariance matrix stored in e(V).
lrplot works with any program as long as Stata's time series operators are used (see tsvarlist), and the program stores the
coefficients and variance-covariance matrix in e(b) and e(V). lrplot replaces the current dataset with the simulated data and the
estimated effects by time period.
lrplot varlist(max=1) [, time(integer 10) sims(integer 10000) seed(string) level(integer 95) line *]
| Option | Description |
|---|---|
| time(integer 10) | number of times periods to simulate. Default is 10. |
| sims(integer 10000) | number of simulations. Default is 10000. |
| seed(string) | seed for random-number generator. |
| level(integer 95) | significance level for confidence intervals. Default 95%. |
| line | plot cumulative effect as a line graph instead of a bar graph. |
| options for the plot (supports graph twoway options). |
lrplot produces a table and a plot of the effects. It is important to note that the effects estimated by lrplot will differ
from those calculated with the nlcom command following estimation. This is because the effects estimated by lrplot are based
on simulations. This also means that a seed must be set to reproduce results.
1. Estimate an ARDL(1,0) with regress
reg y l.y x
lrplot x
2. Estimate an ARDL(1,1) with xtreg, fe over 50 time periods
xtreg l(0/1).(y x), fe
lrplot x, time(50)
3. Estimate an ARDL with a more complex lag structure and a seed to reproduce results
xtreg y l2.y l3.y x l.x l5.x, fe
lrplot x, seed(3424)
4. Plot the cumulative effect with a line graph instead of a bar graph with a custom x-axis title
xtreg y l1.y l2.y x i.time, fe
lrplot x, time(75) seed(3424) line xtitle("Year")
5. lrplot also supports finite distributed lag models (only lags of x appear in the model)
reghdfe y l(0/4).x, a(id time)
lrplot x, sims(1000)
lrplot can be installed by typing the following in Stata:
net install lrplot, from("https://raw.githubusercontent.com/rthombs/lrplot/main") replace
Ryan P. Thombs
(Boston University)
Contact Me: rthombs@bu.edu