Hi,

I've attempted MI with the mi chained command on my dataset of around 6000 with mainly ordinal and binary variables. I'm generating 40 datasets as missingness is around 30-35% on some variables, and have broken down my model using the add(#) replace and replace options to speed things up (code below). I've tried several times but the variables all still have some, albeit less, missingness:

mi impute chained (pmm , knn(10)) FOVWT1 z_c1 = sex ethnicity income_age11 parents_age11 siblings_age11 wellbeing_age11 selfesteem_age11 z_wellbeing_age14 z_selfesteem_age14 z_MH_score, add (40) replace rseed (2002) dots augment noisily

mi impute chained (logit) overweight carer illness parent_depression_age9mo bullied_age11 bully_age11 parent_battles_age11 parent_comm_age11 parks_age11 = sex ethnicity income_age11 parents_age11 siblings_age11 wellbeing_age11 selfesteem_age11 z_wellbeing_age14 z_selfesteem_age14 z_MH_score, replace rseed (2002) dots augment noisily

mi impute chained (ologit) social_network empl_age11 parental_educ_age11 compared_SEP_age11 afford_more_age11 parent_close_age11 parent_health_age11 siblings_bully_age11 parent_satisfied_age11 argue_friends_age11 IMD_age11 area_safety_age11 = sex ethnicity income_age11 parents_age11 siblings_age11 wellbeing_age11 selfesteem_age11 z_wellbeing_age14 z_selfesteem_age14 z_MH_score, replace rseed (2002) dots augment noisily

mi impute chained (logit) teacher_sex suspended_age11 SENDtr_age11 class_misbehaviour_age11 setting_age11 = sex ethnicity income_age11 parents_age11 siblings_age11 wellbeing_age11 selfesteem_age11 z_wellbeing_age14 z_selfesteem_age14 z_MH_score, replace rseed (2002) dots augment noisily

mi impute chained (ologit) stream_age11 prepared_age11 teacher_years_age11 teacher_years1_age11 class_exclusions_age11 = sex ethnicity income_age11 parents_age11 siblings_age11 wellbeing_age11 selfesteem_age11 z_wellbeing_age14 z_selfesteem_age14 z_MH_score, replace rseed (2002) dots augment noisily

Thanks in advance for any advice!