Noobie - 2 months ago 12

LaTeX Question

Consider the following example

`inds <- c('var1','','var2','')`

model1 <- c(10.2,0.00,0.02,0.3)

model2 <- c(11.2,0.01,0.02,0.023)

df = df=data.frame(inds,model1,model2)

df

inds model1 model2

var1 10.20 11.200

0.00 0.010

var2 0.02 0.020

0.30 0.023

Here you have the output of a custom regression model with coefficients and P-values (I actually can show any other statistics if I need to, say, the standard errors of the coefficients).

There are two variables,

`var1`

`var2`

For instance, in model1,

`var1`

`10.2`

`0.00`

`var2`

`0.02`

`0.30`

Is there a package that handle these (custom) tables automatically and can create a neat Latex table with stars for significance?

Thanks!

Answer

Here is a solution using `texreg`

.

Note that `texreg`

>= 1.36.18 is required (as available from its R-Forge project homepage).

The information you are providing in the data frame (coefs and p-values) could be arranged in arbitrary ways in a data frame. Therefore we need to write code that selects these data from the appropriate places in the data frame and uses them to create a `texreg`

object. As you are requesting a generic (and presumably re-usable) solution, we should wrap the code in a re-usable function. I'll call this function `extractFromDataFrame`

. So here is the function, which extracts the information from the data frame and creates a list of `texreg`

objects for the different models:

```
require("texreg")
extractFromDataFrame <- function (dataFrame) {
coef.row.indices <- seq(1, nrow(dataFrame) - 1, 2)
pval.row.indices <- seq(2, nrow(dataFrame), 2)
texregObjects <- list()
for (i in 2:ncol(dataFrame)) {
coefs <- dataFrame[coef.row.indices, i]
coefnames <- as.character(dataFrame[coef.row.indices, 1])
pvalues <- dataFrame[pval.row.indices, i]
tr <- createTexreg(coef = coefs, coef.names = coefnames, pvalues = pvalues)
texregObjects[i - 1] <- list(tr)
}
return(texregObjects)
}
```

In this function, we first define in which rows of the data frame the coefficients are stored and in which rows the p-values are stored. Then we created an empty list in which we stored the `texreg`

objects. We iterate through all columns but the first as the first one contains only the labels. In each of these model columns, we save the coefficients, their names, and the p-values, and then we hand them over to the `createTexreg`

constructor, which is a function that creates a `texreg`

object for us based on the data. We add the `texreg`

object to the list. In the end, we return the list of `texreg`

objects.

We can now apply the function to any data frame that looks like the one provided in the question, with arbitrary numbers of columns (> 1). In this case, after applying the function to the `df`

object, we may want to print the contents of the list if we want to make sure that we did everything right:

```
tr <- extractFromDataFrame(df)
tr
```

And indeed, the results contain the relevant data:

```
[[1]]
No standard errors were defined for this texreg object.
No decimal places were defined for the GOF statistics.
coef. p
var1 10.20 0.0
var2 0.02 0.3
No GOF block defined.
[[2]]
No standard errors were defined for this texreg object.
No decimal places were defined for the GOF statistics.
coef. p
var1 11.20 0.010
var2 0.02 0.023
No GOF block defined.
```

Now we can simply hand the list of `texreg`

objects over to `screenreg`

, e.g., `screenreg(tr)`

, with the following result:

```
========================
Model 1 Model 2
------------------------
var1 10.20 *** 11.20 *
var2 0.02 0.02 *
========================
*** p < 0.001, ** p < 0.01, * p < 0.05
```

Or to `htmlreg`

for creating an HTML table. Or, as requested in the original question, to `texreg`

for creating a LaTeX table. The output of `texreg(tr, single.row = TRUE)`

looks like this:

```
\begin{table}
\begin{center}
\begin{tabular}{l c c }
\hline
& Model 1 & Model 2 \\
\hline
var1 & $10.20^{***}$ & $11.20^{*}$ \\
var2 & $0.02$ & $0.02^{*}$ \\
\hline
\multicolumn{3}{l}{\scriptsize{$^{***}p<0.001$, $^{**}p<0.01$, $^*p<0.05$}}
\end{tabular}
\caption{Statistical models}
\label{table:coefficients}
\end{center}
\end{table}
```

This solution can be modified to accommodate standard errors, confidence intervals, or goodness-of-fit statistics.

Various `texreg`

arguments can be used to customize the output, including the use of the `booktabs`

package or decimal alignment via `dcolumn`

, for example.

Please note that you should not call your data frame `df`

because that object name is already defined in the `stats`

package.