r2 function when only fixed effects are used

[luispfonseca]

Minor issue.

The r2 function returns an error when applied to a regression that only has fixed effects but no other covariates. I haven't checked this for all the objects returned, but adjusted r-squared is returned in the summary function in that case, so I assume it could also be returned when the r2 function is called.

library(fixest)
library(magrittr)

# create data
df <- data.frame(x = runif(1000))
df$id <- sample(20, 1000, TRUE)
df$y <- runif(1000) + 0.2 * df$x + 0.05 * df$id

# show that r2 works well in the normal case
fixest::feols(y ~ x | id, data = df) %>% 
  summary()
#> OLS estimation, Dep. Var.: y
#> Observations: 1,000 
#> Fixed-effects: id: 20
#> Standard-errors: Clustered (id) 
#>   Estimate Std. Error t value  Pr(>|t|)    
#> x 0.204396   0.020233  10.102 < 2.2e-16 ***
#> ---
#> Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#> Log-likelihood: -151.67   Adj. R2: 0.485 
#>                         R2-Within: 0.04161

fixest::feols(y ~ x | id, data = df) %>% 
  fixest::r2()
#>     sq.cor         r2        ar2        pr2       apr2        wr2       war2 
#> 0.49531459 0.49531459 0.48500437 0.69270986 0.65218985 0.04161231 0.04063337 
#>       wpr2      wapr2 
#> 0.12289423 0.11711136

# show that there is an adjusted r-squared measure even for cases where only fixed-effects are used
fixest::feols(y ~ 1 | id, data = df) %>% 
  summary()
#> OLS estimation, Dep. Var.: y
#> Observations: 1,000 
#> Fixed-effects: id: 20
#> Log-likelihood: -172.92   Adj. R2: 0.46319

# error when calling r2 on this estimation
fixest::feols(y ~ 1 | id, data = df) %>% 
  fixest::r2()
#> Error in data.frame(..., check.names = FALSE): arguments imply differing number of rows: 0, 1000

^{Created on 2020-07-06 by the reprex package (v0.3.0)}

Thank you for the great work in the package.

[lrberge]

Thanks for pointing to this issue, now corrected! (The dev. version is not stable though)