I am trying to organize my results obtained with mlogit for exporting to LaTeX with xtable. However, I am finding it difficult to prepare the results in adjacent columns as often found in academic publications.
In particular, I am having problems in the last step, where equations need to be moved next to each other.
I am presenting an example with a small dataframe and how far I have gotten so far below. If there is an easier way to do this, I would be happy if you let me know.
#--------------------------- Create test data and run model --------------------#
id <- 1:12
color <- factor(rep(c("blue","red","yellow"), each=4))
value1 <- round(rnorm(12)*5,1)
value2 <- round(runif(12),1)
factor1 <- factor(rep(c("A", "B"), 6))
data_sample <- data.frame(id, color, value1, value2, factor1)
# Reshape data
data_sample2 <- mlogit.data(data_sample, choice="color", shape="wide" )
# Run model
mlogit.ds <- mlogit(color ~ 1 | value2 + value1 + factor1, data=data_sample2)
#summary(mlogit.ds)
# Save model summary
mlogit.ds <- summary(mlogit.ds)
#-------------------------- Prepare table -------------------------------#
mlogit_table <- data.frame(mlogit.ds$CoefTable)
mlogit_table <- mlogit_table[c(1,4)] # to keep only estimates and p-values
mlogit_table <- mlogit_table[order(rownames(mlogit_table)),] # to group all equations together
mlogit_table
Estimate Pr...t..
red:(intercept) 2.33034676 0.4653448
red:factor1B 0.13591855 0.9506175
red:value1 0.26639321 0.2072482
red:value2 -5.64821495 0.1956896
yellow:(intercept) 5.32776498 0.1372126
yellow:factor1B -3.30689681 0.2688475
yellow:value1 -0.09929715 0.6394161
yellow:value2 -7.28057244 0.1335184
#------------------------ Desired result ------------------------------#
red p yellow p
intercept -0.5522404 0.7597343 0.50745137 0.7349326
factor1B -0.6573629 0.7289306 -0.08885928 0.9528689
value1 -0.4058873 0.1495544 0.05956548 0.7833022
value2 0.6370185 0.8398007 -1.30156671 0.6051921
I need help with creating a solution which could adapt to different numbers of equations (depending on how many levels the outcome variable has) and different lengths of each equation (depending on the numbers of predictors).
Here's one approach:
# extract data
tab <- summary(mlogit.ds)$CoefTable[, c(1, 4)]
# find values of outcome variable
ind <- sub("^(\\w+):.*", "\\1", rownames(tab))
# create table
mlogit_table <- do.call(cbind, split(as.data.frame(tab), ind))
# change row names
rownames(mlogit_table) <- sub("^(\\w+:)", "", rownames(mlogit_table))
The result:
red.Estimate red.Pr(>|t|) yellow.Estimate yellow.Pr(>|t|)
(intercept) -1.9697934 0.3301242 -4.4497945 0.19866621
value2 5.7087164 0.1550275 8.7793979 0.09833026
value1 -0.0838299 0.8377691 -0.4767750 0.29019742
factor1B -0.3583036 0.8447884 0.1671317 0.94356618