Choose Language Hide Translation Bar

Community Trekker

Joined:

Feb 9, 2017

Interpreting a JMP Logit Output

Hi -- I am new to JMP and confused by the logit output.  I need help interpreting the odds and probability of an international plan user and a non-international plan user churning?   Thanks for your help -- I am really not clear on how to interpret the odds ratio output with the level 1/level 2 info??  Thanks for your help!

1 ACCEPTED SOLUTION

Accepted Solutions

Staff

Joined:

Jun 23, 2011

Solution

Re: Interpreting a JMP Logit Output

You have a binary outcome, Churn = no or Churn = yes. You have at least one predictor, User = international or User = non-international. Your logistic regression will fit the logit( Churn) versus the linear model. (Perhaps there are other predictors.)

JMP assumes that the first level is the target level. The first level is genearlly determined by the alphanumeric sorting, special lists of values (e.g., low, medium, high), or defined by the Value Ordering column property. JMP 13 also includes an option in the launch dialog for nominal logistic regression to specify the target level. I assume in your case that would be Churn = yes.

The odds ratio is the ratio of the odds of the target level to the odds of the other level. They are reported as Unit Odds Ratios for continuous predictors. This is the odds ratio per unit change in the predictor. The are also reported as the Range Odds Ratio. This is the odds ratio for the change in the odds from the minimum to the maximum level of the predictor. Your predictor is categorical so you will only have the Odds Ratio report.

The probabilty of Churn is available from the Prediction Profiler. Clck the red triangle at the top and select Factor Proling > Profiler. The Profiler is described in Help.

Learn it once, use it forever!
2 REPLIES 2

Staff

Joined:

Jun 23, 2011

Solution

Re: Interpreting a JMP Logit Output

You have a binary outcome, Churn = no or Churn = yes. You have at least one predictor, User = international or User = non-international. Your logistic regression will fit the logit( Churn) versus the linear model. (Perhaps there are other predictors.)

JMP assumes that the first level is the target level. The first level is genearlly determined by the alphanumeric sorting, special lists of values (e.g., low, medium, high), or defined by the Value Ordering column property. JMP 13 also includes an option in the launch dialog for nominal logistic regression to specify the target level. I assume in your case that would be Churn = yes.

The odds ratio is the ratio of the odds of the target level to the odds of the other level. They are reported as Unit Odds Ratios for continuous predictors. This is the odds ratio per unit change in the predictor. The are also reported as the Range Odds Ratio. This is the odds ratio for the change in the odds from the minimum to the maximum level of the predictor. Your predictor is categorical so you will only have the Odds Ratio report.

The probabilty of Churn is available from the Prediction Profiler. Clck the red triangle at the top and select Factor Proling > Profiler. The Profiler is described in Help.

Learn it once, use it forever!

Community Trekker

Joined:

Feb 9, 2017