Dear all,
I am an epidemiologist, and I will need some help to do survival models because of missing values. This is a retrospective approach based on past collected data from different governmental programs (inconsistency in data sampling). I have a dataset with several mercury exposure for participants at different times (value and month of sampling), their year of birth/death and some sociodemographic covariate that do not vary over time.
I tried a time span approach. The value column time end it the year mercury was measured. (sampling started in 1970)
Ex. Id 1 born in 1960, first data value in 1978
None of the participants had a sampling of mercury the year of their death, then I put a missing value as value. I suppose it is not the good thing. The issues death=1 is always related to a missing value of mercury.
My objective is to do a cox model to test whether Hg is related to death along other covariates non time varying (ex. Sex). Also I was thinking of using “Age sampling_end” instead of “time end” as I would like to do a survival curve with a x axis on age of survival..
stset age_sampling_end, failure(Event==1) id(id_) time0(age_sampling_start). Thank you for helping me. Aline .
I am an epidemiologist, and I will need some help to do survival models because of missing values. This is a retrospective approach based on past collected data from different governmental programs (inconsistency in data sampling). I have a dataset with several mercury exposure for participants at different times (value and month of sampling), their year of birth/death and some sociodemographic covariate that do not vary over time.
I tried a time span approach. The value column time end it the year mercury was measured. (sampling started in 1970)
Ex. Id 1 born in 1960, first data value in 1978
| id | Time start | Time end | Value | Age sampling_start | Age sampling_end | Death status |
| 1 | 1978 | 1978 | 5.6 | 18 | 18 | 0 |
| 1 | 1978 | 1983 | 4.3 | 18 | 23 | 0 |
| 1 | 1983 | 1990 | 3.2 | 23 | 30 | 0 |
| 1 | 1990 | 1997 | . | 30 | 37 | 1 |
| 2 | 1980 | 1980 | 7.6 | 30 | 30 | 0 |
| 2 | 1980 | 1985 | 2.6 | 30 | 35 | 0 |
| Etc. | 1985 | 1996 | . | 35 | 46 | 1 |
My objective is to do a cox model to test whether Hg is related to death along other covariates non time varying (ex. Sex). Also I was thinking of using “Age sampling_end” instead of “time end” as I would like to do a survival curve with a x axis on age of survival..
stset age_sampling_end, failure(Event==1) id(id_) time0(age_sampling_start). Thank you for helping me. Aline .
