Did value ordering change (or break) in JMP 16?

[#1]

Hello!  I've been teaching logistic regression with the Universal Bank data set for years but am getting different results this semester.  The target variable, Personal Loan, has value ordering on it to specify that "Yes" is the class of interest.  In the past, this was necessary to get the parameter estimates to show up relative to the "Yes" class.  For instance, Income has a positive relationship to loan acceptance so it should have a positive coefficient.  However, this semester, the parameter estimates are being displayed relative to the "No" class, which seems backwards compared to the value ordering.  If I remove the value ordering property from Personal Loan, then the model results show up relative to the "Yes" class.  Did something change in JMP 16?  My specific version is JMP® Pro 16.1.0 (539038).  Below are some screen shots of a simple example to show what's happening.  Thanks!

Value ordering on the target variable:

Relationship of Income to Personal Loan:

Simple model fit with Value Ordering:

Simple model fit after removing Value Ordering:

[#2]

The Value Ordering changed in JMP 15.  The sample data table you provided, appears to be a Version 16 copy of the data.  Therefore, it is not appropriate to test under JMP 14.  The Custome Ordering from JMP 16 can not be handled in JMP 14.

[#3 · Accepted solution]

In JMP 16, the default for Target Level was changed to be the high level. Since you have the ordering set as Yes, No in the Value Order property, JMP considers No to be the high level. You can switch them, or remove the Value Order property altogether to make Yes the Target Level.

More appropriately, you can just use the new Target Level column property (added in JMP 16) to specify the level of interest in binary logistic regression models.

[#4]

Thank you!  Very helpful!