Perhaps you've already gotten useful off-list replies.
I assume you can work backwards with the function [-ln(-ln(x))]. (This function, NLOGLOG, looks to be a variant of the ComplentaryLOGLOG function.) It looks like what you compute when you do that is the predicted probability. So, with a 0-1 category variable, a '1' increases the probability of the outcome by an amount 'B'.
Gene Maguin
-----Original Message-----
From: SPSSX(r) Discussion [mailto:
[hidden email]] On Behalf Of Tim Pickles
Sent: Wednesday, February 27, 2013 10:53 AM
To:
[hidden email]
Subject: Negative log log link function in ordinal regression
Hi,
I have recently been running some ordinal regression analyses in SPSS, using the negative log log link function (as the lower categories of the dependent variable are more probable). I have some estimates from this analysis and would like to transform them into odds ratios with confidence intervals.
I am aware that this process would be easier is I was using a logit link function as I could just exponentiate the estimate given, but I am unsure what to do with estimates calculated via a negative log log link function.
Any information/tips/hints on this would be very very useful.
Kind Regards,
Tim.
=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD
=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
Locked
