Dear Statalist Gurus,
I am hoping that someone can offer some advice and show some example code for a problem I am working on. I am using Stata 15, and have read many posts on propensity matching for survival analyses, but have hit some roadbloacks. The nature of my question revolves around multiple treatment groups and survival endpoints.
I am not a statistician, but have spent most of my professional career doing survival outcome analyses. Although I understand how to create propensity scores with Stata in datasets of treated vs untreated individuals, and then perform matching, I have not been able to understand or exploit the tools or packages to do this in datasets with multiple treatments (as opposed to control vs treated).
I suspect what I am asking is simple for a statistician, but hard for me.
I am uploading a sample dataset, but will describe the variables here:
Survival_yrs: number of years from treatment to death or censor event
os_censor: 0= censored, 1=dead
Tx: categorical variable describing the treatment received (Treatment A through E)
Age: continuous variable of Age in years at the start of treatment
CatVar1: categorical variable of 3 (factors or groups) that can alter the risk of death
CatVar2: categorical variable of 5 (factors or groups) that can alter the risk of death
The treatments are in "Tx",
The covariates other than "Tx" that can also alter survival are "Age", "CatVar1" and "Catvar2",
the survival time is "survival_yrs" and death or censor is "os-censor"
Here are my specific questions:
Many thanks to any or all of you willing to help.
Sincerely,
Jonathan Tward
University of Utah
I am hoping that someone can offer some advice and show some example code for a problem I am working on. I am using Stata 15, and have read many posts on propensity matching for survival analyses, but have hit some roadbloacks. The nature of my question revolves around multiple treatment groups and survival endpoints.
I am not a statistician, but have spent most of my professional career doing survival outcome analyses. Although I understand how to create propensity scores with Stata in datasets of treated vs untreated individuals, and then perform matching, I have not been able to understand or exploit the tools or packages to do this in datasets with multiple treatments (as opposed to control vs treated).
I suspect what I am asking is simple for a statistician, but hard for me.
I am uploading a sample dataset, but will describe the variables here:
Survival_yrs: number of years from treatment to death or censor event
os_censor: 0= censored, 1=dead
Tx: categorical variable describing the treatment received (Treatment A through E)
Age: continuous variable of Age in years at the start of treatment
CatVar1: categorical variable of 3 (factors or groups) that can alter the risk of death
CatVar2: categorical variable of 5 (factors or groups) that can alter the risk of death
The treatments are in "Tx",
The covariates other than "Tx" that can also alter survival are "Age", "CatVar1" and "Catvar2",
the survival time is "survival_yrs" and death or censor is "os-censor"
Here are my specific questions:
- How do I generate propensity scores for the 5 different treatment options (variable Tx)?
- How do I generate inverse probability of treatment weights for these 5 treatment options (variable Tx)?
- How do I perform propensity matching when there are 5 treatment groups?
- How do I create psuedopopulations based on the IPTW that I can use to plot adjusted survival curves?
Many thanks to any or all of you willing to help.
Sincerely,
Jonathan Tward
University of Utah