embedding matrix functions into dplyr mutate - r

Question

I have a large dataframe, alldata, for which I'm attempting to do a number of calculations that were previously done on an older version and written in base R. My goal is to create new columns with the outputs of these calculations using dplyr. The previous version of this code used dozens of intermediate dataframes and wrote these calculations out into separate files using functions.
I am curious if it is possible to preserve these functions as they are written, but embed them in dplyr to reference columns in alldata instead of these temporary matrices from the original version.
Here is a sample section of the code I'm working on. As you can see in the comments, I've translated the old R code into dplyr for the simple weighted mean function.
d_weighted = alldata %>%
# equivalent to by = list(regspp = data$regspp[inds]) from old code
group_by(regspp, year) %>%
# equivalent to wgtmean = function(x, na.rm=FALSE) wtd.mean(x=x[,1], weights=x[,2], na.rm=na.rm) from old code
mutate(lat_wgtmean = wtd.mean(x=lat, weights=wtcpue, na.rm=FALSE))
However, as the functions get more complicated than wgtmean, I'd like to just include the function as is.
The next function from the old code calculates a weighted standard deviation from a matrix where the first column consists of values and the second of weights:
wgtsd = function(mat, ...){
x = mat[,1][mat[,2]>0] # trim to values with weight > 0
w = mat[,2][mat[,2]>0]
sqrt(wtd.var(x=x, weights=w, ...))
}
Is it possible to embed this function in dplyr::mutate with value = lat (latitude) and weight = wtcpue (catch per unit effort, converted to weights) to create a new column in alldata that contains weighted standard deviations?
I recognize I could rewrite these functions, but I'd rather not do so for the more complex ones later in the text (see the below one as an example), and I'm curious if there is an elegant solution to integrate functions with matrix arguments with dplyr.
wgtskew = function(mat, na.rm=FALSE){
x = mat[,1][mat[,2]>0] # trim to values with weight > 0
w = mat[,2][mat[,2]>0]
if(na.rm){
s = !is.na(x+w)
x = x[s]
w = w[s]
}
n = length(x)
w = n * w / sum(w) # normalize
if(n>2){
c3 = n / ((n - 1) * (n - 2))
sdv = wgtsd(cbind(x, w), normwt = TRUE, na.rm = na.rm)
xbar = wtd.mean(x, w, na.rm = na.rm)
sk = c3 * sum(w ^ (3 / 2) * ((x - xbar) / sdv) ^ 3)
return(sk)
} else {
return(NA)
}
}

Answer 1

The packages matrixStats and Weighted.Desc.Stat contain many of the functions you might need.
Then you can either Find and Replace using your text editor, or, for example:
wgtsd <- function(...) matrixStats::weightedSd(...)
and run your script as before.