Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

Creating a table with individual trials from a frequency table in R (inverse of table function)

I have a frequency table of data in a data.frame in R listing factor levels and counts of successes and failures. I would like to turn it from frequency table into a list of events - i.e. the opposite of the "table" command. Specifically, I would like to turn this:

factor.A factor.B success.count fail.count
-------- -------- ------------- ----------
 0        1        0             2
 1        1        2             1

into this:

factor.A factor.B result 
-------- -------- -------
 0        1        0
 0        1        0
 1        1        1
 1        1        1
 1        1        0

It seems to me that reshape ought to do this, or even some obscure base function that I have not heard of, but I've had no luck. Even repeating individual rows of a data.frame is tricky - how do you pass a variable number of arguments to rbind?

Tips?

Background: Why? Because it it easier to cross-validate logistic fits to such a data set than the aggregated binomial data.

I'm analysing my with a generalised linear model as binomial regression in R and would like to cross validate to control regularisation of my data since my purpose is predictive.

However, as far as I can tell, the default cross validation routines in R are not great for binomial data, simply skipping entire rows of the frequency table, rather than trials individually. This means that lightly and heavily sampled factor combinations have the same weight in my cost function, which is inappropriate for my data.

like image 215
dan mackinlay Avatar asked Apr 02 '14 20:04

dan mackinlay


1 Answers

You may try this:

# create 'result' vector
# repeat 1s and 0s the number of times given in the respective 'count' column
result <- rep(rep(c(1, 0), nrow(df)), unlist(df[ , c("success.count", "fail.count")]))

# repeat each row in df the number of times given by the sum of 'count' columns
data.frame(df[rep(1:nrow(df), rowSums(df[ , c("success.count", "fail.count")]) ), c("factor.A", "factor.B")], result)

#     factor.A factor.B result
# 1          0        1      0
# 1.1        0        1      0
# 2          1        1      1
# 2.1        1        1      1
# 2.2        1        1      0
like image 187
Henrik Avatar answered Sep 23 '22 03:09

Henrik