• Davor Ivanković and TiljakMirjana KujundzićMK.
• Skola narodnog zdravlja "Andrija Stampar", Medicinski fakultet Sveucilista u Zagrebu, Zagreb, Hrvatska. davor.ivankovic@snz.hr
• Acta Med Croatica. 2006 Jan 1; 60 Suppl 1: 37-46.

AbstractIn the study of the difference between the two or more data groups, first a scientific hypothesis is to be presented involving guessing, questing, and hypothesizing that motivate research. Statistical hypothesis is drawn from the scientific hypothesis, namely, the hypothesis of the researcher (which, as a rule, is affirmative). The mode of statistical hypothesis is presented so as to be valued by statistical and analytical procedures. Testing of the hypothesis is a statistical procedure that can determine whether and how reliably the available data support the given hypothesis. Testing of hypotheses, namely, testing of the significance is basically the procedure of the quantification of impressions regarding the specific hypothesis. The sequence of actions in the testing of the hypothesis: stating a null and an alternative hypothesis; the choice of the significance level (alpha); collection of relevant data based on the adequate sample of subjects; evaluation of the value of results of the statistical test specific for the null hypothesis (Ho); comparison of results of the statistical test with the values from the known value distribution specific for the given test; interpretation of the statistical test results in the probability terms (P-value). The Ho is a supposition of the absence of effect, i.e. that there is no difference between the samples in the population of interest (for example, that there is no difference in arithmetical means). This is a hypothesis of no difference. It is (mostly) made with the aim of rejection. It is either rejected or accepted. The choice of the corresponding statistical test depends on the researcher's design (dependent or independent study design of two or more data samples) as well as on the nature of data (distribution normality and variance homogeneity). Parametric tests carry more strength (the power to detect the difference if in reality it really exists) than nonparametric or distribution free tests. The latter are practical for smaller samples and situations in which the conditions for conducting parametric tests have not been satisfied.

9 keywords...

hide…