On the (mis)use of statistical methods in hydro-climatological research

This paper discusses four areas where, in the author's view, statistical methods are misused in reporting results of hydro-climatological research. They are: (i) the use of the same data set both to suggest a hypothesis (commonly of trend over time) and to test it; (ii) failure to use an appropriate significance level for tests in which a number of hypotheses are tested, even when the data sets used are mutually uncorrelated; (iii) failure to account for spatial correlations between variables, whether these are explanatory or response variables; and (iv) exaggerated importance given to statistical tests of significance, in particular to the 5% and 1% significance levels.