1
$\begingroup$

I am doing multi level analysis with SPSS and have some categorical variables as control variables (country, gender). enter image description here

As you can see country has a significant effect, but I am not sure how I can report it sufficiently with country 3 being set to zero because it is redundant.

Basically I want to make a table which includes Estimates, Standard error and significance. Can I somehow put country as one variable?

$\endgroup$
  • $\begingroup$ Check footnote b... SPSS adds the reference class to the output (which is represented by the intercept and is thus redundant, i.e. getting an effect of 0). $\endgroup$ – Michael M Jul 13 '17 at 10:36
2
$\begingroup$

Redendant levels are most often called reference levels you can just replace 0's in Estimate column with reference level note or change description of non-refrence levels e.g. [country=1 vs 3] or make another footnote under your table saying "Reference levels: country=3, gender=1".

Basically I want to make a table which includes Estimates, Standard error and significance. Can I somehow put country as one variable?

You can get significance of country as one variable. You can't get Estimate and Std. error for it. Why? Just take a look on Estimate's interpretation. 1.07 for country=1 means that in this country OC_A_Total is on average 1.07 units higher than in reference country (country=3), adjusting for age, gender and the rest of your variables, of course. Similarly, you can say, that in country=2 it is 1.005 units higher. These two estimates can not be collapsed into one (at least if you still wish it to have some clear interpretation).

$\endgroup$
0
$\begingroup$

If you really want to avoid this warning, which is due to the dummy variable trap, you can try re-running your model without intercept.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.