I use the plm function of R to run various panel data models. To include time effects for the first difference model I use
model<- plm(y ~ x1 + x2+factor(year) , data = df, model="fd")
R automatically drops the first and the last time dummy. Is it possible to tell R which time dummies to throw? In my case, I need to drop the first and second one. Is there a way to do that?
see this example
df <- data.frame(Country=c('A', 'A', 'A', 'A','A', 'B', 'B', 'B','B', 'B','C', 'C', 'C','C', 'C', 'D', 'D', 'D','D', 'D'),
y=c(12, 15, 19, 19, 29, 12,59, 59, 9,60, 1,3,7,9,3,44,66,77,88,0),
x=c(1+12*2, 1+15*3,1+ 19*1, 1+90*3, 1+29*34, 1+12*8,1+59, 1+59, 1+9,1+60, 1+1,1+3,1+7,1+9,1+3,1+44,1+66,1+77,1+88,1+0),
year=c(1990, 1991,1992,1993,1994,1990, 1991,1992,1993,1994,1990, 1991,1992,1993,1994,1990, 1991,1992,1993,1994 ))
df$id<- rep(1:4, each=5)
Country y x year id
1 A 12 25 1990 1
2 A 15 46 1991 1
3 A 19 20 1992 1
4 A 19 271 1993 1
5 A 29 987 1994 1
6 B 12 97 1990 2
7 B 59 60 1991 2
8 B 59 60 1992 2
9 B 9 10 1993 2
10 B 60 61 1994 2
11 C 1 2 1990 3
12 C 3 4 1991 3
13 C 7 8 1992 3
14 C 9 10 1993 3
15 C 3 4 1994 3
16 D 44 45 1990 4
17 D 66 67 1991 4
18 D 77 78 1992 4
19 D 88 89 1993 4
20 D 0 1 1994 4
mydata <- pdata.frame(df, index=c("Country", "year"))
model<- plm(y ~ x +factor(year) , data = mydata, model="fd")
summary(model)
Oneway (individual) effect First-Difference Model
Call:
plm(formula = y ~ x + factor(year), data = mydata, model = "fd")
Balanced Panel: n = 4, T = 5, N = 20
Observations used in estimation: 16
Residuals:
Min. 1st Qu. Median 3rd Qu. Max.
-65.23957 -13.70019 -0.28284 12.62938 65.88940
Coefficients: (1 dropped because of singularities)
Estimate Std. Error t-value Pr(>|t|)
(Intercept) -1.691089 8.819438 -0.1917 0.8514
x 0.056626 0.049116 1.1529 0.2734
factor(year)1991 20.077837 14.768075 1.3595 0.2012
factor(year)1992 26.674648 17.650245 1.5113 0.1589
factor(year)1993 16.086244 15.558256 1.0339 0.3234
Total Sum of Squares: 15932
Residual Sum of Squares: 12394
R-Squared: 0.22209
Adj. R-Squared: -0.06079
F-statistic: 0.7851 on 4 and 11 DF, p-value: 0.55817
I want to have in the estimation output the time effects for the years 1992,1993, 1994
Simply, adjust levels in your factor variable to the desired order, then run the plm model:
factor (notice year and year_f align: 1990 => level 5, 1991 => level 1, 1992 => level 2...)
df$year_f <- factor(df$year, c("1991", "1992", "1993", "1994", "1990"))
str(df)
# 'data.frame': 20 obs. of 6 variables:
# $ Country: chr "A" "A" "A" "A" ...
# $ y : num 12 15 19 19 29 12 59 59 9 60 ...
# $ x : num 25 46 20 271 987 97 60 60 10 61 ...
# $ year : num 1990 1991 1992 1993 1994 ...
# $ id : int 1 1 1 1 1 2 2 2 2 2 ...
# $ year_f : Factor w/ 5 levels "1991","1992",..: 5 1 2 3 4 5 1 2 3 4 ...
Models (both variants return exact results)
plm + index
model <- plm(y ~ x + year_f, data = df, index = c("Country", "year"), model="fd")
summary(model)
# Oneway (individual) effect First-Difference Model
#
# Call:
# plm(formula = y ~ x + year_f, data = df, model = "fd", index = c("Country", "year"))
#
# Balanced Panel: n = 4, T = 5, N = 20
# Observations used in estimation: 16
#
# Residuals:
# Min. 1st Qu. Median 3rd Qu. Max.
# -65.23957 -13.70019 -0.28284 12.62938 65.88940
#
# Coefficients: (1 dropped because of singularities)
# Estimate Std. Error t-value Pr(>|t|)
# (Intercept) 18.386748 16.783437 1.0955 0.2967
# x 0.056626 0.049116 1.1529 0.2734
# year_f1992 -13.481026 23.736104 -0.5680 0.5815
# year_f1993 -44.147268 41.174228 -1.0722 0.3066
# year_f1994 -80.311349 59.072302 -1.3595 0.2012
#
# Total Sum of Squares: 15932
# Residual Sum of Squares: 12394
# R-Squared: 0.22209
# Adj. R-Squared: -0.06079
# F-statistic: 0.7851 on 4 and 11 DF, p-value: 0.55817
pdata.frame + plm
mydata <- pdata.frame(df, index=c("Country", "year"))
model <- plm(y ~ x + year_f, data = mydata, model="fd")
summary(model)
# Oneway (individual) effect First-Difference Model
#
# Call:
# plm(formula = y ~ x + year_f, data = mydata, model = "fd")
#
# Balanced Panel: n = 4, T = 5, N = 20
# Observations used in estimation: 16
#
# Residuals:
# Min. 1st Qu. Median 3rd Qu. Max.
# -65.23957 -13.70019 -0.28284 12.62938 65.88940
#
# Coefficients: (1 dropped because of singularities)
# Estimate Std. Error t-value Pr(>|t|)
# (Intercept) 18.386748 16.783437 1.0955 0.2967
# x 0.056626 0.049116 1.1529 0.2734
# year_f1992 -13.481026 23.736104 -0.5680 0.5815
# year_f1993 -44.147268 41.174228 -1.0722 0.3066
# year_f1994 -80.311349 59.072302 -1.3595 0.2012
#
# Total Sum of Squares: 15932
# Residual Sum of Squares: 12394
# R-Squared: 0.22209
# Adj. R-Squared: -0.06079
# F-statistic: 0.7851 on 4 and 11 DF, p-value: 0.55817