Warning in `ds.glmerSLMA`: "number of items to replace is not a multiple of replacement length"

[villegar]

 ⚠️ TL;DR

ds.glmerSLMA emits a harmless warning:

"number of items to replace is not a multiple of replacement length"

when different studies return different numbers of coefficients.

The function completes successfully; the warning comes from internal matrix assignment (betamatrix[,k] <- ...).
Padding shorter vectors with NA resolves it.

---

Description

While running the unit tests for dsBaseClient (e.g., test-smk-ds.glmerSLMA.R), the following warnings are observed:

Warning ('test-smk-ds.glmerSLMA.R:106:5'): check slope formulae - 1
number of items to replace is not a multiple of replacement length
Backtrace:
 1. dsBaseClient::ds.glmerSLMA(...) at test-smk-ds.glmerSLMA.R:106:5
Warning ('test-smk-ds.glmerSLMA.R:178:5'): check slope formulae - 2
number of items to replace is not a multiple of replacement length
Backtrace:
 1. dsBaseClient::ds.glmerSLMA(...) at test-smk-ds.glmerSLMA.R:178:5

These warnings occur even when the function completes successfully, and seems to originate from the internal assignment of coefficients into matrices:

• Line ~220–224 in ds.glmerSLMA:

betamatrix <- matrix(NA, nrow=numcoefficients, ncol=numstudies)
sematrix  <- matrix(NA, nrow=numcoefficients, ncol=numstudies)

for(k in study.include.in.analysis){
  betamatrix[,k] <- study.summary[[k]]$coefficients[,1]
  sematrix[,k]  <- study.summary[[k]]$coefficients[,2]
}

The warning is triggered when different studies return different numbers of coefficients (e.g., due to aliasing or missing covariates).

---

Minimal Reproducible Example (Plain R)

The warning can be reproduced without a DataSHIELD server:

# Simulate study.summary as returned internally by ds.glmerSLMA
# study 1 has 3 coefficients
# study 2 has 2 coefficients → shorter than max
study.summary <- list(
  list(coefficients = matrix(c(1, 2, 3), ncol = 1)), 
  list(coefficients = matrix(c(4, 5), ncol = 1))
)

numstudies <- length(study.summary)
numcoefficients.max <- max(sapply(study.summary, function(s) nrow(s$coefficients)))

betamatrix <- matrix(NA, nrow = numcoefficients.max, ncol = numstudies)

for(k in 1:numstudies) {
  betamatrix[, k] <- study.summary[[k]]$coefficients[,1]
}

⚠ R Warning:

number of items to replace is not a multiple of replacement length

---

Suggested Minimal Fix

Replace the direct assignment with a length-safe assignment:

for(k in seq_len(numstudies)){
  ncoef_k <- length(study.summary[[k]]$coefficients[,1])
  betamatrix[1:ncoef_k, k] <- study.summary[[k]]$coefficients[,1]
  sematrix[1:ncoef_k, k]  <- study.summary[[k]]$coefficients[,2]
}

This ensures that shorter coefficient vectors are padded with NA rather than triggering a warning.

---

Additional Notes

• The warning appears when different studies have different numbers of coefficients (e.g., random slopes, missing covariates, aliasing).
• The warning does not indicate incorrect results — it’s only due to R’s column assignment behavior.
• This affects both betamatrix and sematrix inside ds.glmerSLMA.
• This also explains the test warnings seen in test-smk-ds.glmerSLMA.R on lines 106 and 178.

---