Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Fixes future prediction and other bugs #19

Merged
merged 14 commits into from
Dec 12, 2023
Merged

Fixes future prediction and other bugs #19

merged 14 commits into from
Dec 12, 2023

Conversation

petersen-f
Copy link

This PR mainly fixes the prediction and adds options to make sure user can always predict the future as well as minor bugs:

  • adds warning/error when covariates/factors but no training indicator is provided
  • adds option for periodical prediction (same like prophet module) where future data frame for prediction is created that extends beyond provided data + warning that frequency and units should be the same as training data. this way user can predict future without having to provide training indicator/include in training variable
  • forecast verification plot now also works with only one prediction model
  • available models that can be computed are now based on qml logic and not R function which reduces bugs -> e.g. when no covariates are selected, user can only select autoregressive models
  • fixes some minor things dependency issues and qml options as well as enabling weekly seasonality for the prophet module

petersen-f and others added 8 commits November 7, 2023 15:49
@petersen-f
removed parallel processing
60866d8
@petersen-f
added imputation of missing values
7a0d061
@petersen-f
fixed bunch of stuff
f326639
@petersen-f
first attempt at fixing prediction
c3e1cdf 
- added error message when covariates but no training indicator was added
- added periodical based prediction of no training indicator and no covariates are selected
@petersen-f
fixed eBMA prediction adjustment
a390749 
- bma result object was indexed wrong and caused error
- additionally prediction plot now possible with one model only
@petersen-f
fixed prediction, fixed warnings and qml menu
b77d10a 
This commit is a bit large. But in summary the following things have been changed fixed:
- warning of no training indicator is adjusted so that it doesn't show when verification only is selected
- available models are not picked via R function anymore but instead through JS function -> this makes them appear immediately and fixed some bugs with them disappearing
-  fixed bug where BMA predictions were not plotted for forecast verification
- added note to future predictionplot that warns user to select the proper period unit and frequency
- fixed some qml dependencies for enabling future plots and bma
- fixed bug where prophet was unable to detect weekly seasonality (i disabled it for some reason in the past)
@petersen-f
fixes future prediction
d2fad88 
fixes reporting mode message, depenencies and adds warning message for periodical prediction, removes non-working featEng options
@petersen-f
added beta tag and better description
471d10b
petersen-f and others added 5 commits November 8, 2023 21:30
@petersen-f
fixed bug with reopening JASP files
5152029 
- previously the selected prediction models disappeared in-between saves and reopening of the jasp file
- by calculating the available/possible models in an invisible availableVariableList, and then using this as a source for another variable list, the models stay selected
- i also added months to the periodical prediction
@petersen-f
Merge branch 'main' into fix_prediction
5c52d1a
@petersen-f
fixed forecast verification bug and added invariance & lag warning
5b2d412 
- when training window/data has no variance, error is added as bsts and lmSpike package don't work then
- added error when added lagged dependent variable values exceed training window
- fixed error with evaluation plan function and plot slice order
@petersen-f
added warning when training indicator is set wrong
b768038
@petersen-f
fixed lag qml settings and single prediction plot
4855f88
@petersen-f
Copy link
Author

petersen-f commented Nov 30, 2023 •
edited
Loading

This PR now closes issue #2336,#2390,#2455,#2456,#2457,#2458.

It adds quite a few warnings when inputs do not make sense, limits some of the qml options to sensible values and fixes some bugs that prevented computation.

Could you review/approve this PR @vandenman?

@petersen-f petersen-f requested a review from vandenman November 30, 2023 12:38
vandenman
vandenman requested changes Nov 30, 2023
View reviewed changes
R/predictiveAnalytics.R Outdated Show resolved Hide resolved
R/predictiveAnalytics.R Outdated Show resolved Hide resolved
R/predictiveAnalytics.R Outdated Show resolved Hide resolved
R/predictiveAnalytics.R Outdated Show resolved Hide resolved
@petersen-f
Copy link
Author

@vandenman Thank you Don! I indeed found more examples where I did this wrong... But I fixed every gettext now.

@petersen-f petersen-f requested a review from vandenman December 1, 2023 12:17
@vandenman vandenman merged commit 7d84440 into jasp-stats:main Dec 12, 2023
5 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@vandenman vandenman vandenman approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@petersen-f @vandenman