I am currently study the association between PM 2.5 and adult ANCA asthma ( a disease).
Higher PM2.5 is known associated with a higher incidence of respiratory disease.
So, it is possible that higher PM2.5 relates to a higher incidence of ANCA asthma. That is the research question.
I collected pm2.5 data over 5 years, along with patients records to run a historical cohort study using Cox Regression. Codes below:
Code:
stset exitdate, id(uid) failure(ancapos) enter(time admitdate) exit(time exitdate) origin(time admitdate) scale(1)
Code:
stcox pm25 i.male i.admitseason weight comor* avg* ,base
I graphed PM 2.5 value along with time and noticed daily PM2.5 was decreasing with years (e.g. avg PM 2.5 was high in 2016, but low in 2020). Therefore I wonder if the "overall" decline in PM2.5 leads to an negative association with ANCA incidence.(e.g PM2.5 goes lower, ANCA goes higher). I am partially believe that the trend in PM2.5 has interfered with the true association between PM 2.5 and ANCA
I used the log form of PM 2.5 value to remove the long-term trend and seasonality, but nothing changed.
I tried tssmooth shwinter but it shows "backed up" after 500 iterations.
I adjusted seasonality by adding a i.season variable to the model, still , useless.
Not sure what the next step can do to get rid of the trend effect. Any suggestions would be appreciated.
Thank you.
Mikochi
0 Response to Time Trend in PM2.5 and How to Deal with It in Cox Regression?
Post a Comment