Dilemma of null hypothesis in ecological hypothesis's experiment test.

Abstract

Experimental test is one of the major test methods of ecological hypothesis, though there are many arguments due to null hypothesis. Quinn and Dunham (1983) analyzed the hypothesis deduction model from Platt (1964) and thus stated that there is no null hypothesis in ecology that can be strictly tested by experiments. Fisher's falsificationism and Neyman-Pearson (N-P)'s non-decisivity inhibit statistical null hypothesis from being strictly tested. Moreover, since the null hypothesis H0(α=1, β=0) and alternative hypothesis H1'(α'=1, β'=0) in ecological progresses are diffe-rent from classic physics, the ecological null hypothesis can neither be strictly tested experimentally. These dilemmas of null hypothesis could be relieved via the reduction of P value, careful selection of null hypothesis, non-centralization of non-null hypothesis, and two-tailed test. However, the statistical null hypothesis significance testing (NHST) should not to be equivalent to the causality logistical test in ecological hypothesis. Hence, the findings and conclusions about methodological studies and experimental tests based on NHST are not always logically reliable.