multiple imputation chained equations

[abdelilah arredouani]

Dear all,
i am trying to impute some 75 continuous variables with 500 observations using mi impute command. the code i use is below. my naive question is how long it will take. performing chained iterations has been running for more than 30 minutes now. is that normal? or it has to do with the capacity of my computer ( i am using a macbok pro, 2.6 GHz intel core i5, 8gb RAM)

mi set flong
mi register imputed lnbodyfatbmi lnwaisthipratio lnsalivaryamylaseuml lnplasmasalivaryamyalseul lntotalamyalseul lnpancreaticamylaseul lnalbumin lnalkalinephosphatasealp lnaltgpt lnastgot lncpeptideofinsulin lncalcium lnchloride lncholesterol lncorrectedcalcium lncreatinekinase lnestradiol lnferritin lnfolateserum lnfreethyroxine lnfreetriiodothyronine lnggt lnglucose lnhba1c lnhco3 lnhdl lnhomocysteine lninsulin lniron lnldl lnmagnesium lnphosphorus lnpotasium lnserumcreatinine lnSHBG lnsodium lntestosterone lnTSH lntotalbilirubin lnTIBC lntotalprotein lntriglycerides lnureanitrogen lnuricacid lnvitd lnvitaminb12 lnbodyfatbmr lnbodyfatfat lnbodyfatfatmass lnbodyfatffm lnbodyfatheight lnbodyfatleftarmfat lnbodyfatleftarmfatmass lnbodyfatleftarmffm lnbodyfatleftarm lnbodyfatleftlegfat lnbodyfatleftlegfatmass lnbodyfatleftlegffm lnbodyfatleftleg lnbodyfatrightarmfat lnbodyfatrightarmfatmass lnbodyfatrightarmffm lnbodyfatrightarm lnbodyfatrightlegfat lnbodyfatrightlegfatmass lnbodyfatrightlegffm lnbodyfatrightleg lnbodyfattbw lnbodyfattrunkfat lnbodyfattrunkfatmass lnbodyfattrunkffm lnbodyfatweight lnbodyfatwholebody lnwaist lnhip lndiastolicbp lnsystolicbp lnpulse
mi describe
mi impute chained (regress) lnbodyfatbmi lnwaisthipratio lnsalivaryamylaseuml lnplasmasalivaryamyalseul lntotalamyalseul lnpancreaticamylaseul lnalbumin lnalkalinephosphatasealp lnaltgpt lnastgot lncpeptideofinsulin lncalcium lnchloride lncholesterol lncorrectedcalcium lncreatinekinase lnestradiol lnferritin lnfolateserum lnfreethyroxine lnfreetriiodothyronine lnggt lnglucose lnhba1c lnhco3 lnhdl lnhomocysteine lninsulin lniron lnldl lnmagnesium lnphosphorus lnpotasium lnserumcreatinine lnSHBG lnsodium lntestosterone lnTSH lntotalbilirubin lnTIBC lntotalprotein lntriglycerides lnureanitrogen lnuricacid lnvitd lnvitaminb12 lnbodyfatbmr lnbodyfatfat lnbodyfatfatmass lnbodyfatffm lnbodyfatheight lnbodyfatleftarmfat lnbodyfatleftarmfatmass lnbodyfatleftarmffm lnbodyfatleftarm lnbodyfatleftlegfat lnbodyfatleftlegfatmass lnbodyfatleftlegffm lnbodyfatleftleg lnbodyfatrightarmfat lnbodyfatrightarmfatmass lnbodyfatrightarmffm lnbodyfatrightarm lnbodyfatrightlegfat lnbodyfatrightlegfatmass lnbodyfatrightlegffm lnbodyfatrightleg lnbodyfattbw lnbodyfattrunkfat lnbodyfattrunkfatmass lnbodyfattrunkffm lnbodyfatweight lnbodyfatwholebody lnwaist lnhip lndiastolicbp lnsystolicbp lnpulse =  lnage lnheight  lnfastingdurationinminutes lnsalivaryamylas_newuml, add(250) rseed(100)
note: missing-value pattern is monotone; no iteration performed

thanks

[Marcos Almeida]

It seems most (if not all) variables are "not categorical", hence you should prefer - mi impute mvn - command instead of using the chained equations.

[Leonardo Guizzetti]

As for run time, I don't know that we can properly say whether it is too long or not long enough. We have few details of your dataset, and you are generating 250 datasets which may well require a lot of time because of the number of regression models required to be fit.