Question

I’m building 3d contingency tables from 3 variables in a data frame. Let’s suppose I’m constructing these via

table(x,y,z)

Where z is the variable on which I’m stratifying. I’d like to get rid of any (,,z(i)) where the number of observations in that stratum is 1.

How might I do this? I had trouble figuring out how to count observations in the first place, which I thought I’d be able to use, with subset, to pare down my contingency tables.

Answer 1

Supposing your data is contained in a data frame object named data, this code should remove all data in strata with one observation.

data <- data[-which(data$z %in% which(table(data$z)==1)),]

EDIT

This appears to work now. I’m not sure if this will work in general, but it works for this situation.

data <- read.csv(file='~/Downloads/juveniles2forMax.csv')

data <- data.frame(
  Urban = data$Urban,
  RecidivismPlacement = data$RecidivismPlacement,
  timeinjj = data$timeinjj
)

removeStrata <- function(data, z) {
    data[-which(data[,z] %in% as.numeric(attr(which(table(data[,z])==1),"names"))),]
}

removeStrata(data=data, z='timeinjj')