3.1 ReliabilityIn the reliability exercise, the testretest reliability technique was used. According to this method the same instrument is applied to the same respondents at a later stage and the correlation between the two scores is then calculated (Huysamen 1980:54; Mulder 1981:211). The questionnaire was administered by the respective people participating in the exercise. All of these participants are trained in Discus and how to administer the instrument. The instrument was administered for the first exercise to obtain a pretest score. The exercise was then repeated with the same respondents after a period of three months in order to obtain a posttest score. A statistical evaluation of the raw data, resulting from the exercise, was then obtained by using the SAS system, reflecting Pearson's Productmoment correlation coefficient (coefficiency of stability). 3.1.1 The QuestionnaireThe questionnaire consists of 24 questions each of which presents the respondents with four options. The respondents' task is to select one of the options that most closely resembles themselves, and one that least closely describes them. The respondents are required to focus on the role they fulfil in their work environment and answer all the questions in relation to that role.
For the purpose of this exercise the phrasebased questionnaire was used because it is easier to understand. 3.1.2 Sampling Technique and SizeVarious companies were approached to assist with the exercise as reflected in table 1.
The questionnaire was administered to 90 respondents. These respondents were randomly selected from the respective companies reflected in table 1. A statistical evaluation of the raw data resulting from the testing was then obtained by using the SAS system reflecting Pearson's productmoment correlation coefficient. 3.1.3 ResultsThe correlation analyses are reflected in table 2.
The significance level chosen for this instrument is alpha = 5%. Where the pvalue is less than 0.05, the scores show a significant correlation. In the reliability analysis the pvalue in all the cases is as low as 0.0001. This indicates significance at alpha = 0.001. It can therefore be said the correlation is significant at 1% level. The reliability coefficient of the measuring instrument is close to 1 and can therefore be seen as reliable. 3.2 Validity3.2.1 Content ValidityContent validity of the instrument is determined when the instrument is designed. Content validity refers to the extent in which the instrument measures what it is supposed to measure (De Wet, De K Monteith, Steyn & Venter 1981:146; Huysamen 1980:95; Mulder 1989:219). Each question in the Discus instrument was evaluated by the designers of the instrument, namely Axiom. Although content validity was done by Axiom, the researcher also decided to measure validity in terms of criterionrelated validity. 3.2.2 Criterionrelated ValidityBy applying the method of criterionrelated validity, an exploratory study was done by correlating all 15 Factors in Cattell's 16PF with the four dimensions in the Discus. Factor B was not considered as doubts exist in the literature as to the validity of Factor B (intelligence) within a personality test. Criterionrelated validity was restricted to validation procedures in which the test scores of a group of respondents are compared with ratings of other measurements (Aiken 1994:96). Nunally (1978) claims that it is unrealistic to expect exceptionally high correlation coefficients and Anastasi (1976) says that coefficients of 0.20 and higher can be significant. 3.2.3 Sampling Size and TechniqueIn this exercise employees of the Edgars group, Toyota South Africa and Technikon Natal were used. It was therefore decided to use the Discus and Cattell's 16PF (Form A) for this exercise. In an attempt to determine a correlation between the Discus dimensions and Cattell's 16PF, scores on the 16PF were obtained from 120 employees employed by the abovementioned companies. These respondents were randomly selected from line managers, middle managers, professionals and junior officials. The sample was drawn from all organisational functions and cultural groups within the organisations. 3.2.4 Measuring Instruments3.2.4.1 The Discus Questionnaire 3.2.4.2 Cattell's 16PF Cattell applied the technique of factor analysis and obtained a set of 16 primary factors. The rationale behind the 16PF is that a questionnaire which is based on revealed traits, obtained through mathematical techniques from a large pool of possible personality descriptions, is capable of measuring reliably and validly the true constructs present in humans. The general purpose of the 16PF is to describe testees' personality and predict behaviour using a set of selected, structured items. The test has many practical applications, some of which are mentioned below:
For interpretation purposes, factor scores of 1 to 3 and 8 to 10 are considered. The low numbers of each factor are pictured as portraying one extreme of the profile and the high numbers as portraying the other. It should be pointed out that Cattell, in analysing all 16 factors, came up with clusters of several adjectival descriptors for each factor. Table 3 depicts words representative of factor clusters. The KuderRichardson 8method was used to determine the reliability of the 16PF. Table 4 reflects the reliability figure for each factor.
3.2.5 ProcedureThe Discus and the 16PF were administered by trained and registered psychologists. The marking and interpretation of the 16PF questionnaires were done manually by the researcher, a registered psychologist, and by psychologists from the HSRC. 3.2.6 Statistical AnalysisPearson's productmoment correlation coefficient was computed, using the SASsystem to determine a correlation between the identified fifteen 16PF factors and the Discus dimensions. 3.2.7 ResultsThe 120 questionnaires that were returned were statistically analysed. Scores for each of the fifteen 16PF factors as well as the Discus dimensions were correlated. The individual Discus variable scores were then correlated with all the fifteen factors of the 16PF, resulting in significant correlations at 1% level of significance and at 5% level of significance. The results of these findings are reflected in tables 5A (pvalues) and 5B (rvalues).
