Mice Imputation

[khurram_siddiqui]

Hi ,

I am using MICE multiple imputation R package. There is only 879 records out of 14204 missing data which is almost 6% . I started imputing process last night at midnight and now it is 10:00 AM and found it running, it has been almost 10 hours since. How can I boost its performance , having 4 core machine , 16 GB RAM with 64 bit windows 10 OS and 64 bit R is not enough for this imputation using R?

If there is any quicker way to do this job like cloud or anything then please share

Here i am attaching screen of my Machine spec, R version , imputation command and data structure. The data i am using for Big Mart Sales

[andresrcs]

mice() function does not run in parallel so you are effectively working with just one logical processor, take a look at the parlmice() function from the same package for parallelizing this process and taking advantage of all your cores.

https://www.rdocumentation.org/packages/mice/versions/3.4.0/topics/parlmice

Also, I think you are confusing R and RStudio, these are two different pieces of software, the former is the programming language and the later is an Integrated Development Environment. (This is nothing like traditional statistical software like SPSS, minitab, etc.) mice package is written in R but is not an RStudio product, that is why I have changed your topic category to general.

[khurram_siddiqui]

thanks @andresrcs - i am working on it , quick question i have 4 physical core and 8 logical core in my machine. Can i use 6 logical core by considering 8 logical core?

[andresrcs]

Each one of your physical cores can handle two processing threads, so you have 8 logical cores to your disposition.

[khurram_siddiqui]

it means i can use 6 logical cores / thread. Remaining two i want to leave it for other work, am i right?

imp.pmm<- parlmice(data.combined.1,m=5,maxit=10,method="cart",n.core = 6, n.imp.core = 500)

[andresrcs]

Yes, leaving free cores is a good practice to avoid letting your computer unresponsive.

[khurram_siddiqui]

make sense - last question please

using parlmice is not showing up any activity beneath command , mice using without parallel shows some activity as attached in screen. This let me belief that parlmice is not doing anything any thought please?

[andresrcs]

If you can see the icon that says stop, then your R session is busy doing something

Also, you can check your resources monitor and see how many cores are busy

[khurram_siddiqui]

thanks @andresrcs , parlmice is not helping at all its weird simple mice is now running with 5 iterations.