I want to perform a time-dependent Cox analysis with multiple imputations.
Q1. When I perform multiple imputations in conjunction with Cox analysis repeatedly, there were alterations observed
in the outcomes, including changes in the hazard ratio, confidence intervals, and p-values.
Can these alterations be attributed to the implementation of multiple imputations?
(These changes were observed when I put the variables which have missing values.)
Q2. Is this code right for multiple imputations with Time-dependent Cox analysis according to quartile?
Q3. Furthermore, I want to run both baseline Cox and Time-dependent Cox.
Multiple Imputation for missing values and Baseline-Cox:
The results remained the same in both analyses due to the identical follow-up days (end date-baseline date) in the Cox analysis
to the value of differences between the end date and baseline date of the Time-dependent Cox analysis.
In this case, is the same result in both baseline Cox and Time-dependent Cox analysis reasonable?
Your advice would be appreciated.
Q1. When I perform multiple imputations in conjunction with Cox analysis repeatedly, there were alterations observed
in the outcomes, including changes in the hazard ratio, confidence intervals, and p-values.
Can these alterations be attributed to the implementation of multiple imputations?
(These changes were observed when I put the variables which have missing values.)
Q2. Is this code right for multiple imputations with Time-dependent Cox analysis according to quartile?
Code:
mi stset enddate, failure(death==1) origin(baselinedate) id(patientid) mi estimate, hr: stcox ib(1). quartile
Multiple Imputation for missing values and Baseline-Cox:
Code:
mi stset followup, failure(death==1) id(patientid) mi estimate, hr: stcox ib(1). quartile
to the value of differences between the end date and baseline date of the Time-dependent Cox analysis.
In this case, is the same result in both baseline Cox and Time-dependent Cox analysis reasonable?
Your advice would be appreciated.
