Weighted data

When you run statistical tests on weighted tables, the test is always run on the weighted counts. If you want to run the tests on the unweighted data, you must first remove the weighting.

When the table is weighted, you can optionally use a special base called the effective base. This option is selected by default. The effective base is designed to reduce the likelihood of the statistical tests producing significant results because of the adjustments made by weighting; the effective base takes these adjustments into account.

The effective base is also a test of how good the weighting is. If the weighting is inflating the answers from a particular group by a large factor, the effective base tends to be much smaller than both the unweighted and the weighted base. The closer the effective base is to the unweighted base, the better the weighting is.

The effective base is calculated by dividing the squared sum of weights for all of the respondents in the weighting matrix table by the sum of the squared weights.

The Statistics.UseEffectiveBase property controls whether the effective base is used. The property is set to True by default. You can deselect the option by setting the property to False. For example:

TableDoc.Tables.MyTable.Statistics.UseEffectiveBase = False

The option to use the effective base is selected by default. You turn it off by clearing Use Effective Base on the Preferences tab. See Requesting statistical tests for more information.

You can display the effective base on a table by adding an EffectiveBase element to a variable's axis expression. For example:

age{base(), effectivebase(), ..}

You cannot directly enter an expression on the effective base. The calculation of the effective base is based on the preceding base and uses any expression attached to the base. For example, in this axis expression the effective base includes only male respondents, as specified in the base expression:

age{base('gender={Male}'), effectivebase(), ..}

Reference

For an article that describes some methods of adjusting the base to take into account weighting, see Potthoff R., WoodBury M., Manton G. (1992). “Equivalent Sample Size” and “Equivalent Degrees of Freedom” Refinements for Inference Using Survey Weights Under Superpopulation Models, Journal of American Statistical Association, V87, 383-396. This article has an “equivalent sample size” (formula 1.6) that is the same as the effective base.

See also