Izvorni znanstveni članak
ANALYSIS AND INTERPRETATION OF REGRESSIVE SUPPRESSION EXEMPLIFIED ON SOCIOECOLOGICAL ORIENTATIONS
Benjamin Čulig
; Filozofski fakultet, Zagreb
Sažetak
Suppression can be observed very often in cases when an intercorrelation of predictors other than zero exists. Although it is in principle a question of suppression of statistically significant predictors from a multiple equation, in applying this model very complex cases of combined suppression can apear. They make an interpretation of results rather problematic. On an example from the research practice five such cases are being explained. They have been taken from three regression equations, in which a set of predictors containing 24 multivariately normally distributed variables is taken into consideration. These variables have been measured in a value space, while three socioecological orientations have been considered criteria variables. The orientations have been extricated by measuring the various attitudes towards topical problems arising from an individual's relation to nature and technology.
In the first regression there are two typical cases of suppression: a) a variable with an inconsiderable correlation with the criterion, though with a statistically relevant beta ponder; b) two variables with a high correlation and a significant connection with the criterion; however, their beta ponders are not statistically significant. In other two regressions the cases which are combinations of cases as described in a) and b) are being interpreted. Because of the fact that no statistic programme product contains a procedure which would automatically eliminate suppression from a multiple relation, the paper presents an intercorrelation matrix including all regressive variables, by means of which the interpretation of results becomes more meaningfull and methodologically correct. The obtained results suggest an obligatory control of predictor intercorrelation in the course of interpretation, or the application of any of alternative models, especially of those not based on the standard metrics.
Ključne riječi
anthopocentrism; multiple regression analysis; naturalism; socioecological orientations; suppression; technicism; values
Hrčak ID:
141700
URI
Datum izdavanja:
15.9.1999.
Posjeta: 2.742 *