I am running a GEE to test the effect of an ordinal and a nominal independent factor on an ordinal dependent variable. The GEE will use one of each factors conditions as the reference value. For the ordinal value it makes sense to select a descending ordering. However I am not sure how the reference category is determined for the nominal factor. Is it based on the size of the effect ? Is it based on the order in which it appears in the data set? Can you self select the reference category?
Thanks in advance.
Hello Aijse de Vries
The software lets you consider the first or last code as a reference. However, if you want to reference another code, using Syntax is easy.
Abolfazl Ghoodjani Thank you for your quick answer. Can you show me what the syntax is to select the reference category? Keep in mind I need to set it for one of the two independent variables, namely the nominal one.
I have now recoded the variable into a numberred one with labels keeping the meaning clear. Looking whether I can force SPSS to start with the one I labelled as 1 here. Thanks in advance
Aijse de Vries
Look at the syntaxes below.
1- Binary Logistic
LOGISTIC REGRESSION VARIABLES Lung
/METHOD=ENTER Age Health
/CONTRAST (Health)=Indicator(2)
/CRITERIA=PIN(.05) POUT(.10) ITERATE(20) CUT(.5).
2-1 Generalized Linear Models
GENLIN Health (ORDER=ASCENDING) BY Susceptible Practice (ORDER=ASCENDING)
/MODEL Susceptible
DISTRIBUTION=MULTINOMIAL LINK=CUMLOGIT
/CRITERIA METHOD=FISHER(1) SCALE=1 COVB=MODEL MAXITERATIONS=100 MAXSTEPHALVING=5
PCONVERGE=1E-006(ABSOLUTE) SINGULAR=1E-012 ANALYSISTYPE=3(WALD) CILEVEL=95 CITYPE=WALD
LIKELIHOOD=FULL
/MISSING CLASSMISSING=EXCLUDE
/PRINT CPS DESCRIPTIVES MODELINFO FIT SUMMARY SOLUTION.
2-2 Generalized Linear Models
GENLIN Health (ORDER=ASCENDING) BY Susceptible Practice (ORDER=DESCENDING)
/MODEL Susceptible
DISTRIBUTION=MULTINOMIAL LINK=CUMLOGIT
/CRITERIA METHOD=FISHER(1) SCALE=1 COVB=MODEL MAXITERATIONS=100 MAXSTEPHALVING=5
PCONVERGE=1E-006(ABSOLUTE) SINGULAR=1E-012 ANALYSISTYPE=3(WALD) CILEVEL=95 CITYPE=WALD
LIKELIHOOD=FULL
/MISSING CLASSMISSING=EXCLUDE
/PRINT CPS DESCRIPTIVES MODELINFO FIT SUMMARY SOLUTION.
2-3 Generalized Linear Models
GENLIN Health (ORDER=ASCENDING) BY Susceptible Practice (ORDER=DATA)
/MODEL Susceptible
DISTRIBUTION=MULTINOMIAL LINK=CUMLOGIT
/CRITERIA METHOD=FISHER(1) SCALE=1 COVB=MODEL MAXITERATIONS=100 MAXSTEPHALVING=5
PCONVERGE=1E-006(ABSOLUTE) SINGULAR=1E-012 ANALYSISTYPE=3(WALD) CILEVEL=95 CITYPE=WALD
LIKELIHOOD=FULL
/MISSING CLASSMISSING=EXCLUDE
/PRINT CPS DESCRIPTIVES MODELINFO FIT SUMMARY SOLUTION.
Thank you for your reply however this does not solve my problem. As indicated in my question I am aware of specifying the order as descending or ascending. My Question is about how SPSS determines what is the lowest or highest category in a nominal factor.
- Badges
- Science topic