I am using Stata 13.1 for Mac, and iOSX Yosemite.
I am writing to get s steer from members about -mi estimate sctcrreg- after -mi impute-. It seems as if (despite successfully imputing missing data) the estimation is still only using complete cases in the model.
Firstly, the successful imputation
***************start****************
mi impute chained (mlogit) bmi_c alb_c creat_c ....... = age gender PRD freq_cat year_cat CONfreq_c, add(5)
Performing chained iterations .....
------------------------------------------------------------------
| Observations per m
|----------------------------------------------
Variable | Complete Incomplete Imputed | Total
-------------------+-----------------------------------+----------
bmi_c | 6544 2812 2812 | 9356
alb_cat | 8119 1237 1237 | 9356
creat_cat | 5837 3519 3519 | 9356
......
***************finish****************
Following this, -mi stset- also seems successful
***************start****************
mi stset end, id(id) failure(outcome==1) exit(time end) origin(time incept) scale(365.25)
id: id
failure event: outcome == 1
obs. time interval: (end[_n-1], end]
exit on or before: time end
t for analysis: (time-origin)/365.25
origin: time incept
------------------------------------------------------------------------------
9356 total observations
0 exclusions
------------------------------------------------------------------------------
9356 observations remaining, representing
9356 subjects
982 failures in single-failure-per-subject data
29250.81 total analysis time at risk and under observation
at risk from t = 0
earliest observed entry t = 0
last observed exit t = 8.320329
***************finish****************
However, using -mi estimate- seems to use only the complete cases, as evidenced by the "Number of Obs" below?
***************start****************
xi: mi estimate, hr dots: stcrreg age i.gender i.PRD i.bmi_......, compete(outcome = 2 3)
.....
Imputations (5):
..... done
Multiple-imputation estimates Imputations = 5
Competing-risks regression Number of obs = 4204
Average RVI = 0.0000
Largest FMI = 0.0000
DF adjustment: Large sample DF: min = .
avg = .
max = .
Model F test: Equal FMI F( 25, .) = 15.06
Within VCE type: Robust Prob > F = 0.0000
***************finish****************
My questions are:
1) Is stcrreg using only 4204 observations, when 9356 are available? Or am I misinterpreting the "Number of Obs"?
2) If it is only using 4204 observations, why, and how do I get estimates that are combined from modelling in the imputed datasets?
Thanks,
MM
I am writing to get s steer from members about -mi estimate sctcrreg- after -mi impute-. It seems as if (despite successfully imputing missing data) the estimation is still only using complete cases in the model.
Firstly, the successful imputation
***************start****************
mi impute chained (mlogit) bmi_c alb_c creat_c ....... = age gender PRD freq_cat year_cat CONfreq_c, add(5)
Performing chained iterations .....
------------------------------------------------------------------
| Observations per m
|----------------------------------------------
Variable | Complete Incomplete Imputed | Total
-------------------+-----------------------------------+----------
bmi_c | 6544 2812 2812 | 9356
alb_cat | 8119 1237 1237 | 9356
creat_cat | 5837 3519 3519 | 9356
......
***************finish****************
Following this, -mi stset- also seems successful
***************start****************
mi stset end, id(id) failure(outcome==1) exit(time end) origin(time incept) scale(365.25)
id: id
failure event: outcome == 1
obs. time interval: (end[_n-1], end]
exit on or before: time end
t for analysis: (time-origin)/365.25
origin: time incept
------------------------------------------------------------------------------
9356 total observations
0 exclusions
------------------------------------------------------------------------------
9356 observations remaining, representing
9356 subjects
982 failures in single-failure-per-subject data
29250.81 total analysis time at risk and under observation
at risk from t = 0
earliest observed entry t = 0
last observed exit t = 8.320329
***************finish****************
However, using -mi estimate- seems to use only the complete cases, as evidenced by the "Number of Obs" below?
***************start****************
xi: mi estimate, hr dots: stcrreg age i.gender i.PRD i.bmi_......, compete(outcome = 2 3)
.....
Imputations (5):
..... done
Multiple-imputation estimates Imputations = 5
Competing-risks regression Number of obs = 4204
Average RVI = 0.0000
Largest FMI = 0.0000
DF adjustment: Large sample DF: min = .
avg = .
max = .
Model F test: Equal FMI F( 25, .) = 15.06
Within VCE type: Robust Prob > F = 0.0000
***************finish****************
My questions are:
1) Is stcrreg using only 4204 observations, when 9356 are available? Or am I misinterpreting the "Number of Obs"?
2) If it is only using 4204 observations, why, and how do I get estimates that are combined from modelling in the imputed datasets?
Thanks,
MM
