Closed luisvalenzuelar closed 1 year ago
I am afraid, I cannot quite follow. The post you link has an example of a panel FD model estimated by plm
which gives the F-statistic for the joint significance of non-intercept regressors (Wald test) at the end of the printed summary's output:
"## F-statistic: 10.6018 on 2 and 698 DF, p-value: 2.9123e-05"
The minimum reproducible example you give estimates an intercept-only model (ed
is constant within observerational units and exp
gets differenced to all zeros, thus leaving only the intercept). This is why there is no such thing as a Wald test is printed. You can also check via pwaldtest(fd)
for your example which gives the error message "Error in pwaldtest.plm(fd) :
No non-intercept regressors in input model 'x', cannot perform Wald joint significance test".
When you estimate a model with more than just an intercept, the F test is given, see my example below.
I am not sure if I missed anything - could you kindly elaborate?
Your example's printed summery output is:
> summary(fd)
Oneway (individual) effect First-Difference Model
Call:
plm(formula = lwage ~ ed + exp, data = df, effect = "individual",
model = "fd")
Balanced Panel: n = 595, T = 7, N = 4165
Observations used in estimation: 3570
Residuals:
Min. 1st Qu. Median 3rd Qu. Max.
-2.098409 -0.065159 -0.010259 0.049239 2.345911
Coefficients:
Estimate Std. Error t-value Pr(>|t|)
(Intercept) 0.095928 0.003044 31.514 < 0.00000000000000022 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Total Sum of Squares: 118.06
Residual Sum of Squares: 118.06
R-Squared: -0.27826
Adj. R-Squared: -0.27826
Example of a panel FD model with intercept and coefficients:
library(plm)
data("Produc", package = "plm")
zz <- plm(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp,
data = Produc, index = c("state","year"), model = "fd")
summary(zz)
#> Oneway (individual) effect First-Difference Model
#>
#> Call:
#> plm(formula = log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp,
#> data = Produc, model = "fd", index = c("state", "year"))
#>
#> Balanced Panel: n = 48, T = 17, N = 816
#> Observations used in estimation: 768
#>
#> Residuals:
#> Min. 1st Qu. Median 3rd Qu. Max.
#> -0.0852334 -0.0108348 0.0016016 0.0126813 0.1024759
#>
#> Coefficients:
#> Estimate Std. Error t-value Pr(>|t|)
#> (Intercept) 0.01068526 0.00137639 7.7633 2.663e-14 ***
#> log(pcap) -0.00660507 0.04593751 -0.1438 0.8857
#> log(pc) -0.03243575 0.02305050 -1.4072 0.1598
#> log(emp) 0.83147269 0.03696857 22.4913 < 2.2e-16 ***
#> unemp -0.00598593 0.00076141 -7.8616 1.293e-14 ***
#> ---
#> Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#>
#> Total Sum of Squares: 1.0802
#> Residual Sum of Squares: 0.33308
#> R-Squared: 0.69166
#> Adj. R-Squared: 0.69004
#> F-statistic: 427.885 on 4 and 763 DF, p-value: < 2.22e-16
You are right. Problem was data structure. My apologies.
Other models do report the F-Statistic at the end. Apparently, it was reported before (here is a post from 2020). The news file does not seem to show a change on this regard. A MRE below