I am obtaining 2 different results from my data using the sjp.likert command, this is my code:
library(sjPlot)
l <- c(1,2,2,2,3,3,4,4,5,5,5,4,3,2,2)
lab <- c("strongly not agree",
"not agree",
"Neutral",
"Agree",
"Strongly agree")
sjp.likert(items = l,
cat.neutral = 3,
catcount = 4,
legend.labels = lab)
notice that i am working with a numeric variable not factor, at this point everything looks ok, but sometimes i prefer to work with factor to omit the legend.labels parameter. So i use this
l.factor <- factor(x = l,labels = lab)
sjp.likert(items = l.factor,
cat.neutral = 3,
catcount = 4)
But this is where i get the problem, for example: the "neutral" response is not longer 20%, now is 6.7%. As far i can see the package is reading the "neutral" response as neutral, because the grey color in right side.
You can see that the proper number is 20% using this
prop.table(table(l.factor))
prop.table(table(l))
What i am doing wrong? is this a bug?
The problem is that re-ordering the values inside the sjp-likert()-function (if you have a neutral category) is based on numeric indices - which does not work with factors with character levels. So you have to re-order your factor levels before calling the function:
l.factor <- factor(x = l,labels = lab[c(1,2,4,5,3)])
sjp.likert(items = l.factor,
cat.neutral = 5,
catcount = 4)
Another way is to convert factors into numeric values and set the factor levels as label-attribute. You can do this with sjmisc::to_value() with argument keep.labels = TRUE. Taking your example, and modifying it slighlty:
l <- c(1,2,2,2,3,3,4,4,5,5,5,4,3,2,2)
lab <- c("strongly not agree",
"not agree",
"Neutral",
"Agree",
"Strongly agree")
l.factor <- factor(x = l,labels = lab)
l.factor <- to_value(l.factor, keep.labels = T)
sjp.likert(items = l.factor,
cat.neutral = 3,
catcount = 4)
to_value() works both on vectors and data frame, so you can easily convert factors in your data frame into numeric, keeping the value labels:
to_value(my_data_frame, keep.labels = T)