R help - Feb 2025 - Discrepancy in imputed data results across different operating systems

Hi,


I am currently facing an issue with data imputation in R using the random forest
method from the MICE package. Specifically, I have imputed data that was missing
completely at random (MCAR) using the following R code:


# IMPUTATION

set.seed(500)

# Perform multiple imputation

imputed_data <- mice(JPBR, m = 5, method = 'rf', maxit = 50)


Initially, I performed the imputation on a Mac running macOS Big Sur version
11.7.10 with R version 4.2.3.


However, when I attempted to replicate the results using the same seed, data,
and script on a Windows 10 Enterprise machine with R version 4.4.1, the imputed
results differed from those obtained on the Mac.


Concerned by this discrepancy, I ran the same script on another Windows 10
Enterprise system using R version 4.2.3. The results were consistent with the
previous Windows imputation but still differed from the Mac results.


Furthermore, I conducted the imputation on a different Mac running macOS Sonoma
with R version 4.4.1. Interestingly, the results from this setup matched the
original imputation outcome from the Mac, suggesting that the issue may not be
related to the R version but rather to the operating system.


I am seeking insights or suggestions on why these discrepancies occur and how to
ensure consistent imputation results across different operating systems. Any
guidance or advice would be greatly appreciated.


Best,


Johannes


__________________________________

Johannes Bullinger, M.Sc.
Developmental Psychology Department
Ludwig-Maximilians-University of Munich
he/him; er/ihn
Mail: johannes.bullinger at lmu.de


	[[alternative HTML version deleted]]

Normally I'd say you should post a reproducible example where you get a 
discrepancy, but that's already been done in an open bug report here:

https://github.com/amices/mice/issues/688 .

Duncan Murdoch

On 2025-02-20 4:19 a.m., Bullinger, Johannes wrote:
> Hi,
> 
> 
> I am currently facing an issue with data imputation in R using the random
forest method from the MICE package. Specifically, I have imputed data that was
missing completely at random (MCAR) using the following R code:
> 
> 
> # IMPUTATION
> 
> set.seed(500)
> 
> # Perform multiple imputation
> 
> imputed_data <- mice(JPBR, m = 5, method = 'rf', maxit = 50)
> 
> 
> Initially, I performed the imputation on a Mac running macOS Big Sur
version 11.7.10 with R version 4.2.3.
> 
> 
> However, when I attempted to replicate the results using the same seed,
data, and script on a Windows 10 Enterprise machine with R version 4.4.1, the
imputed results differed from those obtained on the Mac.
> 
> 
> Concerned by this discrepancy, I ran the same script on another Windows 10
Enterprise system using R version 4.2.3. The results were consistent with the
previous Windows imputation but still differed from the Mac results.
> 
> 
> Furthermore, I conducted the imputation on a different Mac running macOS
Sonoma with R version 4.4.1. Interestingly, the results from this setup matched
the original imputation outcome from the Mac, suggesting that the issue may not
be related to the R version but rather to the operating system.
> 
> 
> I am seeking insights or suggestions on why these discrepancies occur and
how to ensure consistent imputation results across different operating systems.
Any guidance or advice would be greatly appreciated.
> 
> 
> Best,
> 
> 
> Johannes
> 
> 
> __________________________________
> 
> Johannes Bullinger, M.Sc.
> Developmental Psychology Department
> Ludwig-Maximilians-University of Munich
> he/him; er/ihn
> Mail: johannes.bullinger at lmu.de
> 
> 
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
https://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.