Strange behaviour of categorical predictor in Complex Samples General Linear Model (SPSS)

Question

I'm using the CS-GLM function in SPSS to calculate regression coefficients for a model. CS-GLM is being used because the sample is stratified. The outcome is Somatic Complaint (SC), I have several predictors, one of which is is gender. Gender is at the heart of this problem, because the regression coefficients calculated for it are strange. They are strange regardless of whether gender is the only predictor in a simple regression model or if it's one of many predictors in a multiple regression model. Therefore I'll talk about the simpler model here. In research girls generally experience more somatic complaint than boys. This is shown in descriptive analyses also: girls have a higher mean score of SC than boys. The b-coefficient of gender is negative however. Since gender is coded as boys = 1 and girls = 2, this indicates that boys score higher on SC than girls, contrary to previous research and the descriptive analysis. I'm having trouble figuring out why this happens. This is the syntax used for the model:

CSGLM  SC by GENDER 
  /PLAN FILE='(file path)'
  /MODEL GENDER
  /STATISTICS PARAMETER SE CINTERVAL TTEST
  /PRINT SUMMARY VARIABLEINFO
  /TEST TYPE=F PADJUST=SEQSIDAK
  /MISSING CLASSMISSING=EXCLUDE
  /CRITERIA CILEVEL=95.

If I use WITH instead of BY, indicating that gender is continuous rather than categorical, the b-coefficient that results is positive (which is what I would expect from the descriptive data and previous reserach). But gender is obviously a categorical variable.

Answer 1

I've realised the reason for my confusion: I had misunderstood how to interpret the b coefficients of a categorical predictor. I attempted to interpret them as I would a continuous predictor, instead of a dummy variable. The negative b coefficient indicates that boys experience less somatic complaint when compared to girls. I.e. girls experience more SC than boys. Which is in accordance to both the descriptive analysis and previous research.

A rather basic misunderstanding, but maybe someone else will benefit from it.