library(GGally)
library(corrplot)

let’s load the data, transform variables that should be treated as nominal into factors, and leave the variables that should be treated as ordinal as they are.

bfg <- read.table( "bfg_numeric.csv", sep = "\t", header = TRUE)
bfg$Vote <- as.factor(bfg$Vote)
bfg$Gender <- as.factor(bfg$Gender)
bfg$Race <- as.factor(bfg$Race)
bfg$State <- as.factor(bfg$State)

Reverse code items and re-order the columns so that the items belonging to the same big five dimension are grouped close to each other

ind <- c(2, 6, 21, 31, 12, 27, 37, 8, 18, 23, 43, 9, 24, 34, 35, 41) 
keys <- rep(1, 44)
keys[ind] <- -1
bfg[,1:44] <- as.data.frame(psych::reverse.code(keys, bfg[ ,1:44] , mini = 1, maxi = 5))
ord <- c(1, 6, 11, 16, 21, 26, 31, 36,
         2, 7, 12, 17, 22, 27, 32, 37, 42,
         3, 8, 13, 18, 23, 28, 33, 38, 43,
         4, 9, 14, 19, 24, 29, 34, 39,
         5, 10, 15, 20, 25, 30, 35, 40, 41, 44) 
bfg <- bfg[ , c(ord, 45:57)]

Plot the a subset of data assuming that the ordinal variables are continous

Extraversion <- 1:8
#Agreeableness <- 9:17
#Conscientiousness <- 18:27
#Neuroticism <- 28:34
#Openness <- 35:44
ggpairs(bfg, 
        columns = Extraversion,
        lower = list(
          continuous = 'smooth'
        ),
        upper = list(
          continuous = 'density'
        )
)

Let’s examine some of the remaining variables. The diagonal shows their marginal distribution density. Are they normal? The lower diagonal is a scatter plot of any two pairs of variables with a regression line overlayed on the figure whereas the upper diagonal is a two dimensional bivariate density plot.

ggpairs(bfg, 
        columns = c(45:48, 52:56),
        lower = list(
          continuous = 'smooth'
        ),
        upper = list(
          continuous = 'density'
        )
)

There are some interesting patterns! Now let’s move on to the items.

Since our items are ordinal, we compute the polychoric correlation among the 44 items and plot the correlation matrix heatmap clustered by distance

cor_mat <- psych::polychoric(bfg[ , 1:44])
## 376 cells were adjusted for 0 values using the correction for continuity. Examine your data carefully.
corrplot(cor_mat$rho,
         order = "hclust",
         tl.col = 'black',
         tl.cex=.5)

Notice the top block and how it is negatively correlated with the bigger block, which has 4-5 smaller blocks on the diagonal.