Using texreg (or similar) to produce a regression table for clustered standard error function (cl) output

Question

I've been using the excellent package texreg to produce LaTeX-ready regression tables from fitted model objects, but it doesn't seem to be compatible with various functions that adjust my standard errors for clustering. Some fake data and code gives an example and an error message below.

Any thoughts on how to get the output that I want (similar to what I get from texreg)?

x  = rnorm(1000)
IDs = ceiling(seq(from = .1, to = 10,length.out=1000))
s = c(rep(rnorm(1),100),rep(rnorm(1),100),rep(rnorm(1),100),rep(rnorm(1),100),rep(rnorm(1),100),rep(rnorm(1),100),rep(rnorm(1),100),rep(rnorm(1),100),rep(rnorm(1),100),rep(rnorm(1),100))
y = 3*x + 1.5^(IDs)*rnorm(n=1000,sd=s^2)*x + rnorm(1000)*.3
IDs = as.factor(IDs)
d = data.frame(x,IDs,y)
m = lm(y~IDs+x,data=d)
summary(m)

library(texreg)
texreg(m,omit.coef="IDs")

\begin{table}
\begin{center}
\begin{tabular}{l c }
\hline
            & Model 1 \\
\hline
(Intercept) & $0.12$       \\
            & $(4.50)$     \\
x           & $5.28^{***}$ \\
            & $(1.41)$     \\
\hline
R$^2$       & 0.02         \\
Adj. R$^2$  & 0.01         \\
Num. obs.   & 1000         \\
\hline
\multicolumn{2}{l}{\scriptsize{\textsuperscript{***}$p<0.001$, 
  \textsuperscript{**}$p<0.01$, 
  \textsuperscript{*}$p<0.05$}}
\end{tabular}
\caption{Statistical models}
\label{table:coefficients}
\end{center}
\end{table}

cl   <- function(dat,fm, cluster){
           cluster = as.numeric(cluster)
       attach(dat, warn.conflicts = F)
           library(sandwich)
        library(lmtest)
           M <- length(unique(cluster))
           N <- length(cluster)
           K <- fm$rank
           dfc <- (M/(M-1))*((N-1)/(N-K))
           uj  <- apply(estfun(fm),2, function(x) tapply(x, cluster, sum));
           vcovCL <- dfc*sandwich(fm, meat=crossprod(uj)/N)
           coeftest(fm, vcovCL) }

result = cl(d,m,IDs)
texreg(result,omit.coef="IDs")

Error in (function (classes, fdef, mtable) : unable to find an inherited method for function ‘extract’ for signature ‘"coeftest"’

Did you try following section 5.5 of the package vignette/JSS article (http://cran.r-project.org/web/packages/texreg/vignettes/jss1020.pdf)? — Philip Leifeld

Philip Leifeld Philip Leifeld · Accepted Answer · 2013-10-17T07:10:16

Section 5.5 of the package vignette offers a solution. The package vignette was published as an article in the Journal of Statistical Software. It can be found here: http://www.jstatsoft.org/v55/i08/.

More specifically, the robust standard errors and p values have to be extracted from your result matrix and handed over to texreg via the override.se and override.pval arguments:

se <- result[, 2]
pval <- result[, 4]
screenreg(    # display the results in the R console
    m, 
    omit.coef = "IDs", 
    override.se = se, 
    override.pval = pval
)
texreg(       # for LaTeX output
    m, 
    omit.coef = "IDs", 
    override.se = se, 
    override.pval = pval
)

Another way to do it would be to extract the coefficients from your original model, save them into a texreg object, manipulate this object, and then hand it over to the texreg function:

tr <- extract(m)
tr@pvalues <- result[, 4]
tr@se <- result[, 2]
screenreg(tr)  # or texreg including your original arguments...

Using texreg (or similar) to produce a regression table for clustered standard error function (cl) output

1 Answers