Create a new variable representing the
ratio of each age group in the population to each age group in your sample.
Then weight the file by this new variable. Possible statements:
IF (AGE LT 18)POPWEIGHT=0.28/0.22.
IF (AGE GE 18 AND AGE LE
29)POPWEIGHT=0.14/0.08.
……………………..
IF (AGE GE 65)POPWEIGHT=0.16/0.11.
WEIGHT BY POPWEIGHT.
Of course, the numerator of each ratio is
the true population proportion of each age group, estimated from population
census or other sources; the denominator is the proportion of the same age
group in your sample. My figures are fictitious examples.
In this version, your results will be
still totalizing the same number of cases as your sample size, because the
weights are only proportional. Since the weights are fractional, the resulting
frequencies are rounded, and for subtotals there might be some small
differences from one table to the next for the same frequency, just due to
rounding. Also, totals may not add up exactly for the same reason.
Since sample size is not altered,
significance tests would not be inflated by weighting. Remember that SPSS takes
the WEIGHTED total as “sample size” when performing significance
tests.
Hector
|
Okay, I have conducted a telephone survey of likely
voters. The age of the respondents to the survey came out skewed toward older
respondents. I would like to use SPSS to weight
the results to minimize the skewed effect. Can some one explain to me how to SPSS
to do the weighting? I would like to crate a new variable out of the
weighting procedure to use in the crosstabulation. Thanks
………………………
Ismail K. Noor, Ph.D.,
[hidden email]
http://www.dennonoor.com
(313) 690-0755 (cell)
|
Locked
