<meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />I’m using a logistic regression nominal model for a response variable (pregnant/not-pregnant). My independent variables are: Time of insemination + Production type + interaction. When TOI (48, 52, 56h) is described as a nominal variable both PT and the interaction are significant. When TOI is used is described as a ordinal variable, only the interaction is significant. Why is this relevant to the analysis? I thought the type of logistic regression to be used (nominal or ordinal) was determined only by the data nature of the response variable. Many thanks for any help I can get.
Ian, It took me this long to thank you... sorry. Much appreciated your comments on my question. --Erick
To quote from the 'Statistical Details' section in the 'Fitting Linear Models' document "Factors marked with the ordinal modeling type are coded differently than nominal factors. The parameters estimates are interpreted differently, the tests are different, and the least squares means are different. ".
Take a look at:
http://www.jmp.com/support/help/The_Factor_Models.shtml#93454
for all the details.
