model with three fixed effects in plm package in R

Question

I am trying to estimate the model with 3 fixed effects. One is a customer-fixed effect, another one is good fixed effect and the third one is time-fixed effect. I am new to plm package, but as I understand, if I had just 2 fixed effects (time and good). I would do something like this:

  fe <- plm(outcome ~ dependent variable + explanatory variable 1 + explanatory variable 2, 
            data = mydata, index = c("good_id", "time"), model = 'within', effect = "twoways")

But how I approach this problem in plm package if I have not 2 fixed effects, but 3?

the lfe package allows for higher dimensional fixed effects – paqmo May 06 '20 at 13:56 — paqmo, May 06 '20 at 13:56

score 2 · Accepted Answer · answered May 06 '20 at 10:51

You may add the third fixed effect as a dummy variable using factor(). Example:

library(plm)
data("Produc", package="plm")

# plm FE model
zz1 <- plm(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp + 
             factor(region),
          data=Produc, index=c("state","year"), model='within', effect="twoways")

# LSDV model
zz2 <- lm(log(gsp) ~ 0 + log(pcap) + log(pc) + log(emp) + unemp
         + factor(state) + factor(year) + factor(region),
          data=Produc)

summary(zz1)$coe
#               Estimate  Std. Error   t-value      Pr(>|t|)
# log(pcap) -0.030176057 0.026936544 -1.120265  2.629606e-01
# log(pc)    0.168828035 0.027656339  6.104497  1.655450e-09
# log(emp)   0.769306196 0.028141794 27.336786 1.275556e-114
# unemp     -0.004221093 0.001138837 -3.706493  2.256597e-04

summary(zz2)$coe[1:4,]
#               Estimate  Std. Error   t value      Pr(>|t|)
# log(pcap) -0.030176057 0.026936544 -1.120265  2.629606e-01
# log(pc)    0.168828035 0.027656339  6.104497  1.655450e-09
# log(emp)   0.769306196 0.028141794 27.336786 1.275556e-114
# unemp     -0.004221093 0.001138837 -3.706493  2.256597e-04

Yielding identical coefficients and statistics.

But aren't there any potential problems with transferring it to factor in plm? I haven't fully understood how it can be bad, but from reading this thread (the problem there is different though) it seems that it may cause problems:https://stackoverflow.com/questions/39563595/fixed-effects-in-r-plm-vs-lm-factor Sorry if my question is obvious and there is nothing wrong, I confess that I don't fully get it :) — Anya Pilipentseva, May 06 '20 at 11:17
@AnyaPilipentseva I think your linked answer refers to the error `Error in pdim.default(index[[1]], index[[2]]) : duplicate couples (id-time)`. — jay.sf, May 06 '20 at 11:33
@AnyaPilipentseva Please notice my [new relevant answer](https://stackoverflow.com/a/70845623/6574038). — jay.sf, Jan 25 '22 at 12:58

Anya Pilipentseva · Answer 2 · 2020-05-06T19:15:04.277

0

Thanks to everyone for help! I actually found another solution myself which may add to provided ideas: https://cran.r-project.org/web/packages/fixest/vignettes/fixest_walkthrough.html#1_simple_example_using_trade_data This package allows inclusion of multiple fixed effects

edited May 06 '20 at 19:15

answered May 06 '20 at 14:49

Anya Pilipentseva

117
10

Hi Anya, I know it's been a while since your last comment. Did you finally use this package, or did you use the solution suggested by jay.sf? – Kob May 11 '22 at 06:52
Hi, I uaed fixest package back then and after several times as well – Anya Pilipentseva May 13 '22 at 10:42

model with three fixed effects in plm package in R

2 Answers2