Dear STATA forum,
I have a dataset that looks at whether participants showed up to a visit (yes/no) of which there are 5 visits. I want to know the total number of visits they showed up to and use that totaled variable in Poisson regression. Several of the visits have missing values, and when I add up a participant's visits with missing values, it will result in a missing value. I am struggling with how to apply multiple imputation in this context.
For example, the first 10 rows are shown below corresponding to whether they showed at visits 1 - 5:
I began with the following code:
mi set mlong
mi register imputed visit1 visit2 visit3 visit4 visit5
mi misstable summarize visit1 visit2 visit3 visit4 visit5
mi impute logit visit1 covariate_x covariate_y covariate_z, add(10) rseed(1234)
mi impute logit visit2 covariate_x covariate_y covariate_z, add(10) rseed(1234)
mi impute logit visit3 covariate_x covariate_y covariate_z, add(10) rseed(1234)
mi impute logit visit4 covariate_x covariate_y covariate_z, add(10) rseed(1234)
mi impute logit visit5 covariate_x covariate_y covariate_z, add(10) rseed(1234)
This results in 50 imputations and I see, at this point, there are still several (hundreds of) missing values:
mdesc visit1 visit2
I ultimately want to total the visits and use this variable in my regression model:
gen VisitsAttended = visit1 visit2 visit3 visit4 visit5
glm VisitsAttended varX varY varZ, family(poisson) link(log) eform
Would you total up the visits after performing multiple imputation on each of the visits (as I did)? or is there a more appropriate method to deal with missing values in this scenario?
I really appreciate your thoughts! I am a newbie with MI so apologize if this is basic.
I have a dataset that looks at whether participants showed up to a visit (yes/no) of which there are 5 visits. I want to know the total number of visits they showed up to and use that totaled variable in Poisson regression. Several of the visits have missing values, and when I add up a participant's visits with missing values, it will result in a missing value. I am struggling with how to apply multiple imputation in this context.
For example, the first 10 rows are shown below corresponding to whether they showed at visits 1 - 5:
I began with the following code:
mi set mlong
mi register imputed visit1 visit2 visit3 visit4 visit5
mi misstable summarize visit1 visit2 visit3 visit4 visit5
mi impute logit visit1 covariate_x covariate_y covariate_z, add(10) rseed(1234)
mi impute logit visit2 covariate_x covariate_y covariate_z, add(10) rseed(1234)
mi impute logit visit3 covariate_x covariate_y covariate_z, add(10) rseed(1234)
mi impute logit visit4 covariate_x covariate_y covariate_z, add(10) rseed(1234)
mi impute logit visit5 covariate_x covariate_y covariate_z, add(10) rseed(1234)
This results in 50 imputations and I see, at this point, there are still several (hundreds of) missing values:
mdesc visit1 visit2
I ultimately want to total the visits and use this variable in my regression model:
gen VisitsAttended = visit1 visit2 visit3 visit4 visit5
glm VisitsAttended varX varY varZ, family(poisson) link(log) eform
Would you total up the visits after performing multiple imputation on each of the visits (as I did)? or is there a more appropriate method to deal with missing values in this scenario?
I really appreciate your thoughts! I am a newbie with MI so apologize if this is basic.
