Dear friends,

I am trying to carry out a Multilevel regression analysis with the MIXED procedure. My data file is a EUROBAROMETER with 28 countries, and I applied a weighting variable for adjusting the population of the 28 EU countries (now 27).

When I run the procedure to get the null model, to estimate the intraclass correlation, the MIXED procedure excludes 7 countries from the calculation, the smallest seven countries according to their population. But If I run the procedure without the weighting variable active, then the procedure includes all 28 countries.

¿Please, could someone tell me the reason why the SPSS excludes that seven countries?

The minimum unweighted number of cases in these countries is 500, and 4 of them have 1000 survey observations.

An OCDE methodological guide states that: "As the SPSS® MIXED procedure cannot deal with fractional weights, it is recommended to use SPSS® only for preliminary multilevel analysis. The SPSS® syntax presented in this chapter gives an indication, but will not provide final unbiased estimates".

Is this true?

Thank you very much

Eduardo

From the Command Syntax Reference...

If a WEIGHT variable is specified, its values are used as frequency weights by the MIXED procedure.

• Cases with missing weights or weights less than 0.5 are not used in the analyses.

• The weight values are rounded to the nearest whole numbers before use.

So if the case weights in those countries are less than .5, those cases would be excluded which could result in countries dropping out.