Make calcChoiceProbs more accurate, and make simultaneous evalution f…

[pkofod]

…aster (and more accurate).

The commit message should say it. It makes the simultaneous evaluation faster and more accurate, and the choice prob evaluation more accurate on its own.

Improvements where needed for econ-ark/REMARK#9 to fully replicate DCEGM paper.

[pkofod]

@mnwhite would it be possible for you to look at these changes? It'd probably make sense for them to make it into HARK before the next version, as the way they're calculated here is quite important to be able to replicate the DCEGM paper for example. With the old way of handling overflow, DCEGM figures do not match those published (and the fault is ours)

cc @shaunagm as I feel like this is slightly release blocking

[llorracc]

Yes, I think these should be in the release -- but I thought they'd already been merged into the master branch?

…

On Fri, Apr 5, 2019 at 5:05 AM Patrick Kofod Mogensen < ***@***.***> wrote: @mnwhite <https://github.com/mnwhite> would it be possible for you to look at these changes? It'd probably make sense for them to make it into HARK before the next version, as the way they're calculated here is quite important to be able to replicate the DCEGM paper for example. With the old way of handling overflow, DCEGM figures do not match those published (and the fault is ours) cc @shaunagm <https://github.com/shaunagm> as I feel like this is slightly release blocking — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#242 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABQlf3_h5b-IzCgkzN8dqZ_azqgYiliVks5vdxHvgaJpZM4b76ja> .

-- - Chris Carroll

[pkofod]

Yes, I think these should be in the release -- but I thought they'd already been merged into the master branch?

Some version of the collection of functions are there, but this PR hasn't been merged, so the improvements are not. Edit: so the things we discussed with Shujaat are not in the master branch until this is merged.

[mnwhite]

Ack, I'm in Bordeaux right now, flying to Dublin momentarily. I'm sorry this keeps slipping. I will try to merge in tonight (European).

…

On Fri, Apr 5, 2019, 10:05 AM Patrick Kofod Mogensen < ***@***.***> wrote: @mnwhite <https://github.com/mnwhite> would it be possible for you to look at these changes? It'd probably make sense for them to make it into HARK before the next version, as the way they're calculated here is quite important to be able to replicate the DCEGM paper for example. With the old way of handling overflow, DCEGM figures do not match those published (and the fault is ours) cc @shaunagm <https://github.com/shaunagm> as I feel like this is slightly release blocking — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#242 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ANUQFaCfI8DSscIA2PmJYMly1JGTA_bVks5vdxHwgaJpZM4b76ja> .

[mnwhite]

The loop at 3395 can be done with an indexing scheme (and no loop). But (from my own experience with this exact step) it makes the code harder to read and doesn't actually result in a speedup-- the number of choices is usually like 2-5, so it's not much of a loop to avoid.

Merging now.

[shaunagm]

@pkofod @mnwhite - can one of you suggest a one line summary for the release notes?

[pkofod]

"Improve accuracy and performance of functions for evaluating the integrated value function and conditional choice probabilities for models with extreme value type I taste shocks."