Our World Statistics Day conversations have been a great reminder of how much statistics can inform our lives. Do you have an example of how statistics has made a difference in your life? Share your story with the Community!
Choose Language Hide Translation Bar
Highlighted

Generalized Regression with calculated factors

I want to use the General Regression platform to compare different models.I have several factors, some of which a categorical and others a continuous numeric. I have normalized all factors to the interval [-1;1] and they are all calculated from other columns in the data table.

When I look at the models in the General Regression Platform, I get as expected models based on the calculated factors that I had selected in the model specification window of the Fit Model Platform:

 
 
 

 

image.png

Question: The Model Comparison table at the top of the Generalized Regression Platform indicates for this model "4 nonzero parameters" but I would only expect 3. How are the "nonzero parameters counted"?

 

When I put these models in the Formula Depot (Save Columns -> Publish Prediction Formula) the factors that are listed in the window are the columns that were used to calculate the parameters I use as input for the model fitting (like E^n), not the inputs themselves. (These are listed as "Others".)image.png

Question: Is that the intended behaviour and is there a means to switch that off?

 

If I continue and use start the Model Comparison from the Formula Depot, I obtain the following window. On the top there is a question: "Which column is mean strain (%)_1 predicting?". In the original data table, there is a column mean strain(%) which is used to calculate one of the factors I want to put in the model. (There is no column mean strain (%)_1).

Question: What do I need to do in this window?

image.png

 

For the time being I've selected the column εₘⁿ because that's the factor in the model that is calculated from the column mean strain(%).

 

Choosing the profile in the Model Comparison part of the Formula Depot I get this plot. "Mean strain" appears as output instead of input; what I want to predict is N25. So something seems to be completely wrong.

Question: What am I doing wrong?

image.png

The problem seem to disappear when I copy the calculated factors into a new table and start the model selection from there. So that's a workaround, but I loose the connection to the original table, so that's not an ideal solution.

1 REPLY 1
Highlighted
peng_liu
Staff

Re: Generalized Regression with calculated factors

For the 1st question, Scale is the 4th parameter, i.e. the very last line in your screenshot.

For the rest of questions, both seem to be related to using intermediate formula columns as your predictors with Formula Depot, you may want to contact JMP Technical Support to sort it out.

Article Labels