Differences between version 0.1.1.7 and 0.1.1.9

[ghost]

Hi everyone.

I've used pydlm since version 0.1.1.7 and in fact, in one of my laptops I'm still using it.

I've downloaded the newest version 0.1.1.9 in another lap top and I've checked that results in both versions are totally different (same data, same model).

I've even used the Google unemployed data to check the results in both versions and they'r totally different also. It happens since the model.fit() function and obviously the predict function gives very very strange results in version 0.1.1.9.

I've noticed that version 0.1.1.7 shows results that seem more realistic.

Why does it happen? Did something change in the way modelling data between both versions?

[wwrechard]

Hi Alex,

Without the actual code I cannot say too much, but the package didn't change on API on the trend(). In 0.1.1.7, degree in trend() starts from 1, i.e., degree=1 means constant, degree=2 means linear. Starting from 0.1.1.9, we make the degree more meaningful by changing it to start with 0, i.e., degree=0 means constant, degree=1 means linear. You can find the change in the ChangeLog and the user manual.

If the diff you see was not caused by this change, I would appreciate if you can paste some sample code to help debugging? Thanks!

[ghost]

Thanks a lot, I’ll try it right now :D Honestly I didn’t notice that change because I developed my algorithm in 0.1.1.7 version and I used it with no change in 0.1.1.9 T-T Regards from Mexico. From: Sam [mailto:[email protected]] Sent: domingo, 17 de diciembre de 2017 11:15 p. m. To: wwrechard/pydlm Cc: alexkreamas; Author Subject: Re: [wwrechard/pydlm] Differences between version 0.1.1.7 and 0.1.1.9 (#26) Hi Alex, Without the actual code I cannot say too much, but the package didn't change on API on the trend(). In 0.1.1.7, degree in trend() starts from 1, i.e., degree=1 means constant, degree=2 means linear. Starting from 0.1.1.9, we make the degree more meaningful by changing it to start with 0, i.e., degree=0 means constant, degree=1 means linear. You can find the change in the ChangeLog and the user manual. If the diff you see was not caused by this change, I would appreciate if you can paste some sample code to help debugging? Thanks! — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#26 (comment)> , or mute the thread <https://github.com/notifications/unsubscribe-auth/AYADm-g-llrQeaWozZsDAPaRA27yk-XTks5tBfTogaJpZM4REgKc> .