log transformation of link functions

[gowerc]

Need to clarify the requirements here but Francois mentioned that it might be preferable for the dsld and ttg links to be on the log scale when added to the log hazard.

[gowerc]

Using TTG as an example (its the same for DSLD) the current implementation is that we add the untransformed TTG value to the log hazard. That is:

$$ log(h_i(t)) = log(h_0(t)) + \beta^T X_i + \gamma \cdot TGI_i $$

My understanding of @mercifr1's request is that he would prefer the TGI parameter to instead be on the log scale e.g.

$$ log(h_i(t)) = log(h_0(t)) + \beta^T X_i + \gamma \cdot log(TGI_i) $$

Is that right @mercifr1 ?

[mercifr1]

Thanks for pointing this out. I checked, and it seems that most of the TGI-OS JM published so far are using TGI_i, and not log(TGI_i). So, for the time being, I would keep it that way. If there is the inexpensive possibility to have the log-transformed TGI parameter as an option, that might be worth adding though.

[gowerc]

I think it should be reasonably easy to add it as an option so will leave this open