6.2 Alternatives for Null Hypothesis Significance Testing
In the social and behavioral sciences, null hypothesis testing is still the dominant type of statistical inference. For this reason, an introductory text like the current one must discuss null hypothesis significance testing. But it should discuss it thoroughly, so the problems and errors that occur with null hypothesis testing become clear and can be avoided.
The problems with null hypothesis significance testing are increasingly being recognized. Alternatives to null hypothesis significance testing have been developed and are becoming more accepted within the field. In this section, some alternatives are briefly sketched.
6.2.1 Replication
Another approach that builds upon previous results is replication. If we collect new data on variables that are central in prior research and we execute the same analyses, we replicate previous research.
Replication is the surest tool to check results of previous research. Checks do not necessarily serve to expose fraud and mistakes. They tell us whether prior research results still hold at a later time and perhaps in another context. Thus, we can decrease the chance that our previous results derive from an atypical sample. But replication also helps us to develop more general theories and discard theories that apply only to special situations.