Measuring relationship strength in meta analysis

The primary function of the procedures described so far is to help meta-analysis accept or reject the null hypothesis .Until recently,most researchers interested in social theory and the impact of social interventions have been content to simply identify relations that have some explanatory value.The prevalence of this “yes or no” question was partly due to the relativity recent development of the social sciences. Social hypothesis were crudely stated first approximation to the truth .Social researchers rarely asked how potent theories or interventions were for explaining human behaviour or how competing explanations compare with regard to their relative explanatory value .Today, as their theories and interventions are becoming more sophisticated ,social scientists are more often making enquiries about the size of relationship.

Giving further impetus to the “how much?” question is a growing disenchantment with the null hypothesis significance test itself.If an ample number of participants are available if a sensitive research design is employed ,a rejection of the null hypothesis meta-analyses that include combined significance level ,where the power is not guarantee that an important social insight has been achieved.

Finally,when used in applied social research ,the vote count and combined significance-level techniques give no information on whether  the effect of a treatment or the relationship between variables is large or small,important or trivial.For example, if we find the relationship between whether a particular (a) is a male and (b)believes that women share some culpability when a rape occurs is statistically significant and the correlation is r = .01,is this a ered? What if the result is statistically significant and the correlation is r =.30? This example suggests that the “yes or no?” question is not the question of greatest important .Instead, the important question is , “How much does the sex sex of the participant influence beliefs about rape?” The answer might be zero or it might suggest a small or large relationship.The answer to this question could help meta-analysis (and others) make recommendations about how best to construct rape-attitude interventions so they are most effective .Given these questions ,meta analysis would turn to the calculation of average effect sizes .Also,as we shall see shortly ,the null hypothesis question, “ Is the relationship different from zero?” can be answered by placing a confidence interval around the “how much?” estimate ,removing the need for separate null hypothesis significance tests.