Null Hypothesis

Is Null Hypothesis Significance Testing Significant?
Introduction
Null Hypothesis Significance Testing (NHST) has been one of the most commonly used methods of analysing data collected in biological and psychological fields for over 80 years. There are numerous references in the scientific literature that warns of the limitations NHST brings (Anderson et al). Debate on the efficacy of NHST has been prominent in the social science field, while the ecological and biological sciences have lagged somewhat, however, there has been an increase in discussion and awareness (Anderson et al). One trend has been the emphasis on power analysis link to hypothesis testing, while another trend has been to deemphasize NHST in some disciplines. There are a

NHST and the values it produces can help users with identifying a general pattern or transition to the use of more statistical methods that are more appropriate to their experiment. Individuals still seem to confuse what NHST is and what it is not. Many criticisms lie with its use in an unrelated sense, or in data where the values of the NHST do not further the argument or evidence of an experiment. It is important to remember that NHST can be used for certain situations, however, it cannot deviate and become a blanket statistical test for all data and

One interesting problem with NHST as the primary default for data analysis is that almost all H0s are considered false “on a priori grounds” (Johnson 1995). An example by Savage (1957) brings the point across. “Consider the null H0: θ0 = θ1 = θ2 = … θ5, where θ0 is an expected control and the others are ordered treatment responses.” Anderson et al (2000) provides a few more examples of priori false hypotheses: H0: μ1=μ2, H0: SjC=SjD and H0: pyx=0. The null hypotheses are all surely false. This does not allow for experiments and their results ability to advance science (Savage 1957), or give insights for potential conservation, planning, management and future research (Anderson et al 2000). These issues should rather focus on estimation of effects on differences as well as their precision instead of testing a null hypothesis. This is because whether rejected or not, null hypothesis usually provide little to no use (Morrison and Henkel