Tables, Sig Testing & Weighting Question

[Valdiserri, Colin]

Hello everyone,

 

I have a question about using the tables function on a weighted data set
and significance testing. When running tables we typically significance
test in tables, but we are working with a data set that is weighted and
what we need to do is a little different.

 

We need to run tables weighted to get the weighted percentages, but we
need to significance test without the weight. This might sound odd, but
I can explain it more detail if needed.

 

The only way we can figure out how to do this is to run tables weighted
and then manually significance test with the unweighted sample sizes
using an Excel program we wrote.

 

Does anyone have any direction whether or not we can have SPSS Tables to
all of this for us so we do not have to add an additional step of manual
significance testing? So basically we would need Tables to give us the
percentages using the weight and then significance test those
percentages using the unweighted sample sizes.

 

Please let me know if you need further clarification for this question.

 

Thank you,

 

Colin

 


_______________________________________________________________________

The information transmitted is intended only for the person or entity to which it is addressed and may contain confidential and/or privileged material. Any review, retransmission, dissemination or other use of, or taking of any action in reliance upon, this information by persons or entities other than the intended recipient is prohibited. If you received this in error, please contact the sender and delete the material from any computer.
_______________________________________________________________________

[Hector Maletta]

Colin,
when you say testing without weighting you probably mean without inflationary weights that inflate totals to population size. That is correct. If you test the significance of anything, SPSS uses the total weighted cases as its "sample size", and therefore will believe that your sample is as large as the population, leading to regard almost everything as significant.

But doing without inflationary weights does not mean doing without weights. If your weights differ across cases (as they usually do) because cases in the various strata or domains were selected with different probability, then you still need to weight the cases, albeit in a non inflationary way. A typical non inflationary weight is (proportion in population) divided by (proportion in the sample).

Using these weights with SPSS (without the Complex Samples component) could generate unbiased significance tests only if the sample did not use clustering but only stratification. Clustering is the intermediate selection of groups of cases (e.g. cities, or city blocks) within a stratum (e.g. a State), before the final selection of actual cases (e.g. households or persons). Clustering increases sampling error.
Is your sampling model did not include clustering, non inflationary weights could be used and ordinary sig tests with SPSS could give you correct results. Otherwise you must use Complex Samples.

You may want to browse my article on "Weighting" at www.spsstools.net (look for the Tutorials section) where this is discussed more at length.

Hector


----- Mensaje original -----
De: "Valdiserri, Colin" <[hidden email]>
Fecha: Jueves, Marzo 29, 2007 5:06 pm
Asunto: Tables, Sig Testing & Weighting Question

> Hello everyone,
>
>
>
> I have a question about using the tables function on a weighted
> data set
> and significance testing. When running tables we typically
> significancetest in tables, but we are working with a data set
> that is weighted and
> what we need to do is a little different.
>
>
>
> We need to run tables weighted to get the weighted percentages,
> but we
> need to significance test without the weight. This might sound
> odd, but
> I can explain it more detail if needed.
>
>
>
> The only way we can figure out how to do this is to run tables
> weightedand then manually significance test with the unweighted
> sample sizes
> using an Excel program we wrote.
>
>
>
> Does anyone have any direction whether or not we can have SPSS
> Tables to
> all of this for us so we do not have to add an additional step of
> manualsignificance testing? So basically we would need Tables to
> give us the
> percentages using the weight and then significance test those
> percentages using the unweighted sample sizes.
>
>
>
> Please let me know if you need further clarification for this
> question.
>
>
> Thank you,
>
>
>
> Colin
>
>
>
>
> _______________________________________________________________________
>
> The information transmitted is intended only for the person or
> entity to which it is addressed and may contain confidential
> and/or privileged material. Any review, retransmission,
> dissemination or other use of, or taking of any action in reliance
> upon, this information by persons or entities other than the
> intended recipient is prohibited. If you received this in error,
> please contact the sender and delete the material from any computer.
> _______________________________________________________________________
>