Choosing tests of statistical significance in studies of social interactive behaviour: A critical note on murray, hayes, and smith (1978)

Significant chi-square differences for boys and girls on variables of observational period and adult presence-absence, reported by Murray, Hayes, and Smith (1978), were found to be not significant when analysed by the Wilcoxon matched-pairs signed-rank test and t test. The discrepancy between the outcomes of the different analyses is seen to be related to the inappropriateness of chi-square as a test of significance using behavioural data which have been summed across individuals within the sample (Cochran, 1954; Lewis and Burke, 1949). Appropriate ways of using chi-square to analyse social interaction and other methods of analysis, including analysis of variance, information analysis and transition probability, are reviewed.