Study title

Abstract copyright UK Data Service and data collection copyright owner.

PSST has been written for the user who has a set of measures on a sample of cases and wishes to look at the differences or associations between pairs of variables: it caters for the assortment of measures which are common in psychological or sociological research (classifications, ranking scales and interval scales). The user specifies any pair of variables whose relationship he wishes to examine, and the program selects statistical tests according to whether it is the difference or the association between variables which is to be tested, and the nature of the variables in each pair. Significance levels are given, as well as two-way tables (where practicable). The user may also obtain a printout and tabulation of selected variables. `Dud' scores may be catered for, and new variables can be created from the data read in (`transgeneration'). The limits on the number of cases and variables are set only by the amount of core store available on the computer; unlike most packaged programs, this one automatically re-writes its own storage declarations according to the user's requirements. This procedure has the added advantage of ensuring that the absolute minimum of core store is used, which can reduce computing costs.
The statistical tests available are Mann-Whitney's U, Wilcoxon's T, Chi-squared, Kendall's tau, Product- moment Correlation, Goodman and Kruskal's Gamma, and t-tests. In most cases more than one test is used. This allows the user to compare the power of different tests in handling various types of data: the user should not, of course, adopt a policy of simply quoting the most significant statistic on each occasion.