I have a dataset containing 10 emetic episodes with corresponding time of occurrence as well as plasma concentrations collected from two animals following administration of a single dose.
I thought I ´d try to model the data using a repeated time to event model which I ´ve never done before so I ´m wondering how to reset the hazard function after an event? And also, what needs to be included in the censor column?
In the censor column, when you have an event you put 1 and you leave empty if no event. At the end of the experiment (last time point), you put a zero to emphasize right censoring.
Now if you are using the event statement, the cumulative hazard is reset automatically after each observation. Then nothing to do there. Note that the program stops integrating after reaching a PK observation too but the censor column is zero or empty and therefore the likelihood calculated at that time point for the event is the one associated with no event and then you still have a reset. The reason is that you need to reset after stopping at any observation.
event(DV,hazardrate)
You put the expression for the hazard rate.
The program automatically integrates the hazard rate from one observation to the next one and reset the cumulative hazard (not the hazard rate) which is the integral of the hazard rate from the previous observation to the new and observation.
Thank you Serge for your quick reply, much appreciated! However, when I try to run the model I get an error message saying that the launch process failed and that the system cannot find the file specified. I ´ve attached the project in case you have the time to take a look at it.
I strongly recommend you to read the PML Reference Guide from the page 25. I’m not sure that event() is the right way here (are you sure about resetting hazard after each episode?)
And the hazard function shouldn’t be the same as the standard Emax function (please see PML guide) Moreover, hazard with time-dependent variables is very (VERY) tricky