I am trying to fit a cumulative risk line for a binomial regression analysis<meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />I would like to find something in the Fit Y by X function that includes a 95% CI and a predictive interval.
I have been using the Generalized Linear Model function in the fit model menu (with the binomial option), but I can't plot a predictive interval, even though the profiler function gives me a 95% CI. See below.
None of the options in the Fit Y by X menu allow me to make the same plot, which includes the predictive interval option, but no means that I can tell to fit a cumulative risk line.
Any ideas?
Thanks in advance.
Michael
Re: I am trying to fit a cumulative risk line for a binomial regression analysisThanks very much Byron - I will play with this!
Michael
Re: I am trying to fit a cumulative risk line for a binomial regression analysisMichael,
I'm not sure I understand completely, but both the platforms have an option to calculate inverse predictions (red triangle menu). The report includes confidence intervals for each inverse prediction. Make a combined table of your inverse predictions and plot the data in Graph Builder.
Figure1 combines the graph builder figure with the inverse prediction figure. (using right click, Edit/Copy Frame Contents)
-The black arrows are from the Inverse Prediction figure in the Fit Y by X platform, and the colored lines are a smooth fit from Graph Builder.
-The default inverse prediction dialog allows 10 results, so you might have to do multiple inverse predictions to get the number of points you want for your figure.
-Sample Data: Penicillin.jmp
Granted this is kind of an ugly figure, but it shows that the inverse prediction confidence interval data can be graphed using the Graph Builder. Adding more intervals would result in a nicer looking curve. (And it doesn't require manipulating formulas or writing scripts.)
Hope this helps,
Byron
