Self-enhancing Comparisons in Aerobics - Research Paper Example

Critical analysis of the research paper Being better than some but not better than average: Self-enhancing comparisons in aerobics Module: Concepts and Issues in Multivariate Analysis The study in question – “Being better than some but not better than average: Self-enhancing comparisons i n aerobics” – was a social experiment that compared the ways in which students evaluated themselves under a variety of conditions pertinent to the sporting background. The sport being studied was aerobics – a sport that is open to Social Comparison since the activities and accomplishments of any one student are visible to all others in the group. The study attempted to measure the effects of two independent variables - the subjects performance standard, and the standard of the class – on the type of comparison made by the subjects towards particular and generic others. It was hypothesised that: 1. Across all classes, students would show a preference for comparing upwards with a particular other. 2. Subjects in the threat condition would make a downward comparison to a particular other. 3. Across all classes, students would make a downward comparison to a generic other. 4. Students in the threat condition were less likely to make a downward comparison to a generic other. Thus, the effects and interactions of the two independent variables were observed on two dependent variables. The first independent variable – the subjects performance standard – had two levels for the study – high and low. The second independent variable was the performance standard of the class; also divided into two levels – high and low. Four experimental groups were formed: i. Student with high performance in high performance group. (Uniform) ii. Student with high performance in low performance group. (Advantage) iii. Student with low performance in low performance group. (Uniform) iv. Student with low performance in high performance group. (Threat) Each group consisted of 25 of the total 100 participants in the study. This was done to ensure that a majority of the participants represent the test condition. The authors seem to have overlooked the possibility that as ‘new’ individuals; the participants may pay more attention to each other as compared to the rest of the group. Also, the presence of 24 others would dilute the threat and advantage conditions – an effect which is not considered. These participants were recruited under a deception condition, but complete care was taken that no harm – mental or physical – was caused to the participants in the long term. There are nevertheless ethical concerns about the use of a deception condition. The classes from which the participants were recruited were quite heterogeneous in their composition of age, ability and other possibly relevant factors. This heterogeneity would affect the chosen sample as well as their perception of the test condition group. These factors could confound data; as could factors like sex (there were disproportionately more women) and number of participants per condition (which was 25 per cell, and thus, considered a small sample size). There is also a concern about the scoring of the self with a generic other. The participants had only one exposure to the test setting; and thus may have chosen to compare to their regular class (which is more familiar) to the test setting to which they were primed. A variety of statistical procedures were carried out in order to test the hypotheses put forth. The study used Two Way ANOVAs, students t – tests, correlations and the chi square test. The two way ANOVA’s allowed the authors to study the interaction effects between the two independent variables, while the t-test’s helped test for differences between the absolute mid-score and the mean comparison score. Chi – Square was used to test for cross tabs analysis to support the results of the t-tests and the ANOVA. Correlations were used to develop indices of comparison and to test the relationship between the dependent variables. All these tests require the data to be normally distributed, and with the exception of the Chi-Square, they also require that the variances be equal for all groups and the data to be on at least an equal interval scale. There is concern about all these assumptions being met as no pre-testing is mentioned; and the validity of the choice of the tests, as the use of univariate tests where multivariate ones are applicable do increase the possibility of Type I error. As all participants came from a relatively closed community of the University; there is also concern about the generalisablity of results. Initially, all the sets of scores for the particular comparison other were inter-correlated, and since the correlations were quite high, one comparative index was created for all these scores. The Cronbachs Alpha, - an estimate of reliability - showed that the index would have a high reliability (0.89), and could be used without too many concerns. A high Cronbach’s Alpha indicates that each of the items on a test share strong correlations with the total score of all other items on the test, meaning that all items on the said test show consistency in their measurement of a construct. Although internal consistency is a good measure of reliability; a validity analysis would have strengthened the case for the use of the comparative index. A similar index was also computed for the scores on comparison with a generic other. The Chronbachs Alpha was very high (0.90) for this index as well. Such systematic evaluations of reliability are important since the data was collected in rating form on a set of dimensions for each dependent variable. Next, for the actual testing of the hypotheses, two 2 x 2 ANOVAs were computed. The interaction effects were analysed in order to test the hypotheses. Although none of the hypotheses predicted any main effect differences; in order for the hypotheses to be truly verified, one would see significant main effects as well as the interaction effects. The study expected significantly distinct scores in one of the four cells, which would, in theory, contribute towards differences in row and column means across categories. The first 2 by 2 ANOVA was conducted on data about comparison to a particular other. The statistical analysis showed that the main effects for both variables were significant; but the one for standard of class (p < 0.001) was much stronger than the one for the standard of subjects performance (p Read More