I realise there are better forums to ask, but does anyone have experience of multilevel imputation using mice in R

[StealthPolarBear]

And be willing to answer some idiotic questions?

Lunch is nothing!

Magpie it's been running since yesterday evening, still on the first imputation of 100

R still gives me shudders when I think about it

Can you explain what predicottss for each predicted variable are - I'm not familiar with this term and can't find it in mice documentation although only done quick search - exactly what is being calculated?

Also what is your sample size, number of clusters, number of predictor variables, method of imputation for each variable? They all impact on processing time.

Yes, ok, that is a bit long!

Sounds like you have an endless loop and it's not actually doing anything, SPB, try it with just 2 imputations and 2 rows or some such, smaller sample dataset anyway.

I don't think I do. It is making progress just very slowly. Works fine on smaller datasets. I think I'm going to have to stop it

Thanks op, you've just given me a flashback to R! Thankfully some years ago now but doubt I'll ever forget it

I have expertise in this method and fitting these models. If the script has been running that long with no progress, then it's not likely to lead anywhere useful. The model may be poorly specified, or if it is well specified, then computing power is not sufficient to run it efficiently. Problems of this nature can't be solved without details on the size of the dataset, the number and size of clusters, patterns of missingness in variables within and across clusters, the method of imputation used for each variable, the number and structure of predictor variables, and choices regarding other parameters (for example what is the justification for generating 100 imputed datasets?).

I have absolutely no idea what that means

Well I have definitely been known to run stuff overnight but I usually get results back after 12 hours! How big is your dataset?

I am forever running sql scripts overnight!
Anyway, I am pleased to report this morning it is almost done, looks like another hour or so. Then I want to turn it 'long' which I can see how to do, and feed it back into sql server, which I am still working out.

I have no idea what you're talking about, but I read the whole thread and I'm very impressed!

Like the shipping forecast

Exactly! this was a lovely thing to read despite me having no idea of the meaning.
Love seeing MNers’ superpowers in action