Our objective is to describe living conditions based on a representative sample of individuals and households. By "representative", we mean that we have applied probability sampling procedures such that each individual and household in the population has a known chance of being selected.
In order to make comparisons of sub-populations like regional center residents and village residents, we grouped the population (stratified it) and sampled from each group. The probability of selection could differ between groups, but not within groups. For example, the probability of selection could differ between households in regional centers and villages, but not within a regional center or within a sampled village.
Sampling weights take into account known differences in the probabilities of selection between groups to yield a representative sample. There are four sampling weights for the Alaska dataset:
The two weights beginning with "REG" yield estimates for the three regions (N. Slope, Bering Straits, Northwest Arctic) as a whole. The weight including "HH" yields estimates for households. The weight including "IND" yields estimates for Inupiat adults, aged 16 and over.
Similarly, the two weights beginning with "VIL" yield estimates for households ("HH") and adults ("IND"). These weights should be used when comparing results for just the sampled communities.
Note that most tests of statistical significance are sensitive to the size of the sample. In SPSS these tests use the weighted sample size if the sample is weighted and the actual sample size if the sample is unweighted. Since the weighted sample sizes are much larger, tests of significance using weighted data appear to show significant differences when such differences are actually not significant. When comparing weighted data, a rough but useful test of significance can be obtained by running the same comparison on unweighted data.