Abstract
The use of significance tests in social sciences is widespread mainly due to simple computation via statistical packages. Unfortunately the more social scientists use statistical significance estimates for making causal inferences the less they appear to understand about this influential concept.
Statistical modelling results are usually presented in terms of their statistical significance and little other information is provided. The goal of this article is to show the limits of using statistical significance as a sole means of making inferences; and to present alternative statistical fit indicators readily available within frequentist approach to statistics: confidence intervals, minimum sample size and power analysis.
Multiple working hypotheses are also explored together with two well known information criteria – AIC and BIC. This article provides practical information on how to undertake valid and reliable statistical analyses of social science data.