www.phidot.org

Using robust design models, I am getting very similar standard errors for my beta values for parameters where the structures are an interaction of two categorical factors. Below is an excerpt from the MARK output for a model where S(sex*season). I just wanted to put out feelers and see if anyone has experienced anything similar, and/or if anyone sees this as problematic.

Code: Select all

LOGIT Link Function Parameters of { S(~SEX:Season)Gamma''(~SEX)Gamma'(~SEX)pi(~1)p(~session)c(~session) }
                                                              95% Confidence Interval
 Parameter                    Beta         Standard Error      Lower           Upper
 -------------------------  --------------  --------------  --------------  --------------
    1:S:(Intercept)          0.2279715       36.828500      -71.955890       72.411833   
    2:S:SEXF:SeasonFall      0.5941570       36.829831      -71.592313       72.780627   
    3:S:SEXM:SeasonFall      0.9831622       36.830056      -71.203750       73.170074   
    4:S:SEXF:SeasonSpring   -0.2447608       36.828685      -72.428984       71.939462   
    5:S:SEXM:SeasonSpring    0.3811150       36.828570      -71.802884       72.565114   
    6:S:SEXF:SeasonSummer   -0.0984097       36.828839      -72.282935       72.086116   
    7:S:SEXM:SeasonSummer    0.5520083       36.828778      -71.632398       72.736415   
    8:S:SEXF:SeasonWinter   -1.1145898       36.829233      -73.299887       71.070708   
    9:S:SEXM:SeasonWinter   -0.8813907       36.828749      -73.065741       71.302959


[It looks like this was generated with RMark.] If so use -1+sex:season or sex*season. The problem is that you have a model with an intercept and each interaction (combination) which is one more parameter than you have combinations of sex and season.