马赛克图：空单元格和重叠标签

Question

I am using vcd library in R in order to perform a mosaic plot of 2 categorical variables, having multiple levels each. I had to customize the function call "mosaic" in order to solve my problem of overlapping labels.

However, I have 2 questions regarding this function mosaic :

1. Why proper line break for long level names is not done automatically, when the simple call is done ? By looking at various vcd tutorial, it looks the function by default should handle that.
2. Why vcd is plotting empty cells ? For example, in my dataframe study_filt, I have no observation with the homosex variable set to "Other". Can it be removed ?

>

library(dplyr)
library(vcd)
library(MASS)

load(url("http://bit.ly/dasi_gss_data"))

study=dplyr::select(gss,year,homosex,vetyears)
study=filter(study, year>1974)
study=filter(study, year<1995 | year>2009)
study=group_by(study,year)
study_filt = filter(study,!all(is.na(homosex)) && !all(is.na(vetyears)))
study_filt <- as.data.frame(study_filt)

Simple call to the function mosaic

vcd::mosaic(data=study_filt,~homosex+vetyears)

Customized call to handle long level names

vcd::mosaic(data=study_filt,~homosex+vetyears, 
        labeling=labeling_border(
        rot_labels = c(90, 0, 90, 0), 
        just_labels=c("left","left","right","right"),
        tl_varnames = FALSE,
        gp_labels = gpar(fontsize = 9)))

Answer 1

1. I don't think that labeling_border supports automatic line breaks, just rotation and abbreviation as standard strategies for long labels.

2. mosaic() shows all levels(study_filt$homosex) even if they have zero occurrences. This might be an important piece of information in some scenarios. In your case it is easiest to drop the factor level by calling factor() another time: study_filt$homosex <- factor(study_filt$homosex)