ejosymart/sizemat

Estimate Size at Sexual Maturity.

Hi there, I'm Edgar Josymar! 👋

sizemat's Issues

classify_mature() - Function assigns adults as juveniles

Dear Josymar Torrejon-Magallanes,

I am currently trying to undertake some size of maturity analysis on some Whelk data, I am trying to classify the juveniles from adults using the classify_mature() function. However when plotting the results the PCA seems to deem the smaller individuals as Adults and the larger ones of Juveniles. A screen shot is below: (x = total shell length, y = penis length)

the call of classify_mature():

Is there a solution to this issue??

Link to GitHub repos from DESCRIPTION file

It took a while to locate this repos from https://cran.r-project.org/package=sizeMat. In order to easier find this repository, please consider adding a URL (and BugReports) entry to the DESCRIPTION that link to https://github.com/ejosymart/sizeMat/.

Allow manual creation of `classify` objects

It might be useful to allow for the case where maturity has been determined elsewehere (not via classify_mature()) but the size at maturity still needs to be calculated via morph_mature() (guess who had this use case today?).

Prototype:

manual_classify <- function(x, y, mature) {
   mat_data <- data.frame(x = x, y = y, mature = mature)
   class(mat_data)  <- c("classify", class(mat_data))
   return(mat_data)
}

If there's nothing else that I'm missing (in terms of interacting with other functions) then sure, it's simple enough, but appending a specific class to a data.frame isn't something that every user will know how to do (correctly) to satisfy morph_mature()'s requirement.

Apologies if I've missed any existing implementation.

Excessive memory usage and clean-ups

The new_data list in .morph_mature_fq() stores an excessive amount of data. It retains the entire list of niter re-samples despite these only being required in the glm on the line following, within the same loop.

sizeMat/R/morphMat-internal.R

Line 10 in 4d760a9

new_data[[i]] <- data[sample(nrow(data), nrow(data), replace = T), ]

This quickly used up available memory when the input data was large and the default number of iterations were used.

I suggest the following replacement:

set.seed(seed)
n_coef <- list()
for(i in seq_len(niter)) {
  new_data   <- data[sample(nrow(data), replace = TRUE), ]
  model_boot <- glm(mature ~ x, data = new_data, family = binomial(link = "logit"))
  glm_coef   <- coef(model_boot)
  n_coef     <- rbind(glm_coef, n_coef)
}

I also recommend updating 1:niter to seq_len(niter) which better handles the niter = 0 case (not explicitly forbidden, good practice in any case), and using TRUE rather than T. sample by default takes size to be the "number of items inferred from the first argument", which expands to 1:nrow(data), so the second nrow(data) is redundant.