This section deals with statistics and how QPSMR Companion marks significant differences. Please refer to the Overview of statistics page for more information.
If you apply respondent weighting to a table, the program maintains weighted and unweighted counts for all cells and is able to use the “effective sample sizes” when calculating statistics. If you believe your weighting procedures to be relevant and correct, then you can produce meaningful statistics.
The “effective sample size” is the base figure taking into account any “weighting effects”. The term “weighting efficiency” is used to describe the relationship (expressed as a percentage) between the effective sample base and the unweighted base.
You can perform Z tests or t-tests on tables with normal rows using format SIG. Companion tests each row independently.
For tables with score values assigned to the rows it is better to test the mean scores using a t-test. For tables with a value (integer or float) as the rows it is also better to use a t-test on the average value.
If the breakdown (columns on the table) contains a multi-coded section then you can use overlapping t-tests for rows and means. For more information: Statistics formats settings
You can see some of the formulae used by Companion on the Formulae page.
For a description of one tailed tests, please visit Companion’s One tail or two tail page.