search

IDEMSInternational / R-Instat

A statistics software package powered by R
http://r-instat.org/
GNU General Public License v3.0
38 stars 102 forks source link

Design of predict sub-dialog #2246

Open stevekogo opened 7 years ago

stevekogo commented 7 years ago

@rdstern This sub-dalog is needed by the use model dialogues. Could we have a design of how it should look like?

rdstern commented 7 years ago

You are right that we now need to "attack" the use model dialogues. I still don't have a very clear idea of how it will look. I think I know the sort of things I want. The fit model and choose model dialogue (2 variable up to general) will produce regression objects. The use model selector will use those objects.

Part of my confusion is because we are trying to cope with such a wide range of types of data for the y-variable in the single dialogue. I think that is good in a way. But we must be careful we don't pay too high a price, either in the dialogue becoming too complicated for our users, or in the results not being displayed in ways that would only be appropriate for one type of data.

It seems obvious that we will still allow a look at the residuals, but the fitted value graphs become more important. And we will want predict to also give fitted value tables. These can be at the different factor levels for the factors, and at specified example points for variates.

I hope this isn't a diversion, but we also want to display the results from fitted models clearly. I think (at this stage) this probably comes best in our choose model dialogues, but we may want to asses this as we go along. A related point on all these dialogues Fit, Choose, Use (I think) is to display the results in better ways than the default output from lm or glm etc.

I wrote a long time ago - March - about the possibility of using (for example the stargazer package and I would still like to consider this.

I think what is needed is for us to spend some time in RStudio with all these, before trying to get to specific menus. And we need to document carefully, because I am not sure we can "finish" before turning attention (at least briefly) to the climatic dialogues.

After all that, to answer more specifically, your request for a PREDICT sub-dialogue. Here is the predict sub-dialogue from Genstat. That is for glm. We will need to translate many of the components into equivalents for R commands.

genstat predict dialogue

More is to come. I am now getting into this - after looking more into Genstat!

dannyparsons commented 7 years ago

@rdstern this still looks useful, tagging as 0.9.x (post July)