# Comparing logistic coefficients on different logit models?

I've developed a logit model to be applied to six different sets of cross-sectional data. All the models have same independent variables but the sample size is different for each model. What I'm trying to test is the impact of each independent variable on dependent variable across all the models say for example income in model 1 vs. income in model 2 vs. income of model 3 etc. I have observed from the Odds Ratios, that these variables have different impact on dependent variable for different household categories and regions.

My question is: How do I statistically test this change?

Update:

Further information regarding my question: I'm trying to look at the effects of IVs on DV at different disaggregated levels. The models are: $$y_1 = \beta_{01} + \beta_{11}x_1 + \beta_{21}x_2 + \beta_{31}x_3 + ... \\ y_2 = \beta_{02} + \beta_{12}x_1 + \beta_{22}x_2 + \beta_{32}x_3 + ... \\ y_3 = \beta_{03} + \beta_{13}x_1 + \beta_{23}x_2 + \beta_{33}x_3 + ... \\ y_4 = \beta_{04} + \beta_{14}x_1 + \beta_{24}x_2 + \beta_{34}x_3 + ... \\ y_5 = \beta_{05} + \beta_{15}x_1 + \beta_{25}x_2 + \beta_{35}x_3 + ... \\ y_6 = \beta_{06} + \beta_{16}x_1 + \beta_{26}x_2 + \beta_{36}x_3 + ...$$ The question is can I test whether $\beta_{11}$ is statistically different from $\beta_{12}, \beta_{13}, \beta_{14}, \beta_{15}, \text{ & }\beta_{16}$ etc.?

-
So, this is not a problem of model selection, you want to compare effect-size measure across models? –  chl Jul 23 '12 at 7:54
This sounds like meta-analysis (you can also search CV by clicking on this tag -> meta-analysis). Is there any reason why the effect of these IV's might / should differ? –  gung Jul 23 '12 at 13:43
Thanks... I'm trying to look the effects of IVs on DV at different disaggregated levels. The models are: y1= b01 + b11x1 + b21x2 = b31x3 + ... | y2 = b02 + b12x1 +b22x2 + b32x3 +.... | y3 = b03 + b13x1 + b23x2 + b33x3 + ... | y4......... the question is can i test that b11 is statistically different from b12, b13, and so on? –  Muhammad Khalid Bashir Jul 24 '12 at 1:49
You can do this using a seemingly unrelated regression approach. Be careful comparing coefficients across models: the scale parameter is not fixed, so the coefficients are on different scales. Put differently, a variable can have the same marginal effect in two models, but the coefficients can be of different magnitudes. –  Charlie Mar 19 at 22:49