[Importance and limits of statistical methods in psychiatry]

Clinical reasoning and statistical treatment of psychiatric data interact to reduce the risk of random diagnosis and prognostic probabilities. Errors of statistical procedures are found in an important amount of papers in the most reliable psychiatric journals. Guide-lines for reporting statistics are given advising the authors about mistaking statistic application. Unfortunately, on a more elementary level, representation of data are often criticizable for they are incomplete, or make difficulties of decoding. Harmonization of data representation is then desirable. Some remarks are expressed about paucity of negative studies in literature.