truncated distribution with two border??? about bmstdr

[sujit-sahu]

Thanks for the suggestions. Here are my answers:

Does a limit of 0 mean that 0 is modeled in the distribution or only values > 0?
0 is modelled in the sense that P(Y=0) is allowed to have a positive value depending on the fitting data. Obviously, P(Y>0) = 1-P(Y=0) will be set according to the model.

How do I determine the correct/optimal lambda value for the option 'truncation.para'? Currently I use the default setting.

As far as I understand, lambda determines the power transformation. It will be safer to try a few values of lambda and then choose the optimal value according to some validation criteria.

Setting an upper limit for the data, e.g. 0 <= y <= 100.

That is a very interesting suggestion. But I do not think it has been implemented in the spTimer package. I do not see an easy way to implement that either. One needs to write the whole model fitting code from scratch for this.

But I think there can be a way out to model your data still with the existing code. Say you have 0<y<1. Then just apply either the inverse-logit or inverse-probit transformation z=Phi^{-1}(y). Then you can model z which can take any value on the real line.

Please let me know what you think.