Closed kenkellner closed 4 years ago
Looks great. Just a couple of minor requests. Could you let predict
return the entire set of samples from the posterior predictive distribution, instead of just the mean, SE, and CI? Also, for the SE, I think you want to use sd
instead instead var
, right?
I'll look at the docs later.
After you wrap this up, I'd say we're ready for 1.0. What do you think?
I did take sqrt()
of var()
, so I think the values are correct, but I don't know why I didn't just use sd()
.
Do you think the function should return the summary stats AND the samples, or just the samples?
Regarding 1.0 I think we're almost ready. In addition to this I want to fix #164 and #168 first but should get them done this weekend.
If possible could you go through the old issues quick and close any that aren't relevant anymore? There are a few where I can't tell if they're fixed or not, eg #62, #63, #81. I don't think all outstanding issues necessarily need to be addressed before 1.0 but just want to make sure all the important stuff is done.
My bad. I didn't notice that sqrt
.
I suppose it might be easier for users if predict
returned the summary stats and samples.
Sounds like a good plan for 1.0. Thanks again.
predict
now returns all the samples. I put an example of calculating summary stats on the output in the help file, rather than a dedicated function.
User supplies the
ranef
object and a custom function, and predict returns an estimate, SE, and 95% CI for the function's output parameters. Also adds aposteriorSamples
function that just grabs the samples, so you can work on them separately. See example below. Fixes #166.@rbchan, if you could quickly look over the implementation and the docs that would be great since I'm not sure I explained everything correctly.
Source package: unmarked_0.13-2.9003.tar.gz Binary: unmarked_0.13-2.9003.zip