Gini coefficient, ineq package in R and integer overflow

问题

I want to calculate the Gini coefficient for each column in a 2090 x 25 dataframe. I am using the Gini function in the ineq package and the following code:

gini <- sapply(mydata, function(x) ineq(x,type="Gini")).

This produces results that look valid but also the following warning message:

Warning messages:

1: In n * sum(x) : NAs produced by integer overflow
2: In sum(x * 1:n) : Integer overflow - use sum(as.numeric(.))
3: In n * sum(x) : NAs produced by integer overflow

To overcome the integer overflow I converted the dataframe to a matrix (mymatrix <- as.matrix(mydf)) but then the results were all zeros or NAs. I think this is because ineq package requires a vector and matrix is not a vector.

My questions are:

• how can I convert integer columns to numeric and retain a vector class?
• are there any other options to work around the integer overflow problem?

Thanks

Nerida

回答1:

In absence of more info, my guess would be you might prefer

sapply(1:25, function(x) ineq(as.numeric(mydata[,x],type='Gini')) )

Edit: as @James and @Roman pointed out, sapply will grab each element of a dataframe in turn, so

sapply(mydata,function(x) ineq(as.numeric(x),type='Gini') )

should produce the same result.

来源：https://stackoverflow.com/questions/16890686/gini-coefficient-ineq-package-in-r-and-integer-overflow