I wonder whether Kaplan Meier method and Cox PH model can be applied to the pesticide research. If a insect population has resistant and susceptible individuals. Under the stress of pesticide, the resistant individuals that survive longer than the wildtypes are more likely to be censored if experiment time is limited. It seems to me the assumption is violated because censoring is informative.
If the concordance index is less for example c-index 0.5. How we can improve this concordance index in R language? If we used optimisation method then which parameters we need to used in Coxph function. Thank you.
I am ploting the effect of 4 diets conditioned by the presence of infection, therefore two categorical variables but my graphs in kaplan meier are crossing and in all cases cox.zph is less than 0.05, so ssumption of a proportional hazard is violated; what should i do ???
hi, it depends a lot of the goal of you analysis... as there are pros/cons/limitations to the different approaches. one solution is to "stratify" on the variable(s) that violate the PH assumption. a limitation of this is that you dont estimate a coefficient for any variables that you stratify on, so if this is a concern, then this isn't going to be a good approach. the other common approach is to fit a model that allows for "time-dependent parameters/coefficients". this allows the coefficient (and hence the hazard ratio) to vary over time. in concept what this is doing is including an interaction between the variable that violates the PH assumption and time...making the coefficient for that variable change/depend on the time. id suggest reading a bit more about either of these, and that should get you started...
Hi loved your videos. i am facing an issue with kaplan meier graphs. On plotting two population( control and test), the graph crosses. Also in cox, there is an over fitting issue as c-index on training is too perfect and test c-index is low. I think i am facing this due to time varying covariate. How do I deal with this ? I m using lifeline package in Python.
If your graphs in kaplan meier are crossing then your assumption of a proportional hazard is violated and Cox proportional hazards model cannot be used. You might need to look for non proportional hazard models
Amazing job. Survival regression can be so intimidating at first, but this clearly breaks it down.
Amazing video first time ever get to understand CPH model. Please do some more videos on AFT model
Your videos are amazing/clear/informative/helpful. Thank you :)
you're welcome ;)
I wonder whether Kaplan Meier method and Cox PH model can be applied to the pesticide research. If a insect population has resistant and susceptible individuals. Under the stress of pesticide, the resistant individuals that survive longer than the wildtypes are more likely to be censored if experiment time is limited. It seems to me the assumption is violated because censoring is informative.
you videos are excellent Could you please create a class on co-variance structure matrix in generalized linear models
Thanks very much! It's so clear and easy to understand!
where is this video taken from? it's different from the others. Looks like it's from another course?
If the concordance index is less for example c-index 0.5. How we can improve this concordance index in R language? If we used optimisation method then which parameters we need to used in Coxph function. Thank you.
Nicely explained..Thank you
thanks :)
I am ploting the effect of 4 diets conditioned by the presence of infection, therefore two categorical variables but my graphs in kaplan meier are crossing and in all cases cox.zph is less than 0.05, so ssumption of a proportional hazard is violated; what should i do ???
hi, it depends a lot of the goal of you analysis... as there are pros/cons/limitations to the different approaches. one solution is to "stratify" on the variable(s) that violate the PH assumption. a limitation of this is that you dont estimate a coefficient for any variables that you stratify on, so if this is a concern, then this isn't going to be a good approach.
the other common approach is to fit a model that allows for "time-dependent parameters/coefficients". this allows the coefficient (and hence the hazard ratio) to vary over time. in concept what this is doing is including an interaction between the variable that violates the PH assumption and time...making the coefficient for that variable change/depend on the time.
id suggest reading a bit more about either of these, and that should get you started...
Hi loved your videos. i am facing an issue with kaplan meier graphs. On plotting two population( control and test), the graph crosses. Also in cox, there is an over fitting issue as c-index on training is too perfect and test c-index is low. I think i am facing this due to time varying covariate. How do I deal with this ? I m using lifeline package in Python.
If your graphs in kaplan meier are crossing then your assumption of a proportional hazard is violated and Cox proportional hazards model cannot be used. You might need to look for non proportional hazard models
thank you very much
Great 😁
Why do you laugh