Help on interpretation of Logistic Fit Y by X

[#1]

Hello,

I observed this problem a few times already, and kindly need help to interprete my data correctly:

A nominal outcome (recieved a medication: milrinone) is compared to a continuos variable by Fit Y by X (age in years). A logistic fit X is generated:

It appears to me that with higher age the frequency of subjects recieving milrinone is higher.

Quite the opposite is true, if I turn the variables around or use age is an ordinal variable:

The graph in the logistic model should show the same direction as in the ordinal model (less proportional use in higher age), the calculated OR in the logistic model should also be inversed ?

I may completely wrongly interpretating the logistic fit model ?

Or how can I turn it around ?

With many thanks, Marc

[#2]

Look closely at the logistic regression output - it says " For log odds of 0/1."  So, an upward sloping curve means that older ages are associated with a 0 being more likely.  When you run the logistic regression you can tell JMP to reverse the 0/1 order - or just make sure to look at the small print before interpreting the output.

[#3]

Thanks !

But isn't the graph at the logistic fit indicating the opposite (1 is up / 0 down) ?

How could I tell JMP to reverse the order ?

Many thanks !

[#4 · Accepted solution]

The right side of the logistic plot indicates the marginal probabilities. The probanility of 0 is plotted at the tick mark. The probability of 1 is 1-Pr(0). The logistic curve indicates the conditional probabilities. The probability of 0 is the curve. The probability of 1 is 1-Curve. An upward sloping curve means that the probability of 0 outcome is increasing.

JMP uses the natural ordering. The first level therefore is 0 and the second level is 1. JMP always computes the log odds for first level versus second level. You can either use different values or apply the Value Ordering column property to impose the order that you want.

[#5]

Great ! Thanks !