# quota or stratified sampling?

Discussion in 'SPSS' started by wollyka, Nov 29, 2011.

1. ### wollykaGuest

Hello, Everyone.

I need your help, I have a question regarding quota, stratified
sampling and area sampling. so it is not directly SPSS related
If a quota has been picked including key variables such as region,age
and gender. Then the researcher went to each region and start
selecting households randomly in that region and filling
questionnaires with individuals to fill the quota.

The sampling method here is still quota sampling? or is it a mixed of
stratified and quota sampling?

Also, since quota sampling is regarded as stratified sampling with a
more or less non-random selection of units within strata, sampling
error formulas cannot be applied to the results of quota samples. And
I gather we should not apply the stratified random mean calculations
to results obtained through quota sampling, or can we?

How do we weight the results when there is now an element of
randomness in the quota sampling? That is to say, the selection of
respondents is done randomly until the pre-set quotas are reached.

I'd love to hear what you think please.

wollyka, Nov 29, 2011

2. ### Jon PeckGuest

This is a multi-stage sample. You can define each stage using the Complex Samples wizard either for drawing a sample or preparing for analysis. Thenthe complex samples procedures will use that sample definition to calculate the appropriate standard errors etc.

Jon Peck, Nov 30, 2011

3. ### wollykaGuest

Thanks

But theoretically, (without using SPSS) , it is called quota or multi-
stage sampling?

wollyka, Nov 30, 2011
4. ### JonGuest

Well,
Strictly speaking Jon's reply is not entirely correct. It may be
correct that it is a multi stage sample (assuming that there is random
selection of regions, if there is not, then it is a stratified quota
sample. But, you cannot consider the quotas as clusters). However,
there are two problems with using the complex samples option. The
first is that you do not know what the inclusions probabilities are,
although you could fake it by using for example census data for age
and gender (many will simply assume equal probabilities). The second
is that it is, after all, a quota sample, and the variance estimation
is not designed to handle that. There is no obvious way for
calculating standard errors. That filling the quotas is a more or less
random process within each quota is fairly common. If the selection is
truly random, and from a list, one could have considered it a
stratified sample, but again, the inclusion probabilities pose a
problem.
best
jon

Jon, Dec 9, 2011