VAUS, David de. 2002. Analyzing Social Science Data. London: Sage. ■Mi = n ** >j. v> í 9 r^s í fí ií n --. S 9 I ESI* ö 55 S^ř -■- II I RH ľ ■3 c -— oo STK oS 1! íl u ; ,_ H st Ü Bš i> I« «i zťkš ■- •' H3 t - r> .-r -a o-o fl> ej p oj « o Q P trs - — 9 t- ° 2-5 5: rí-~: S-í WS-B É. S - -. - ros ■a» Si 5 «5 - 3 a- — O. o. Sk K! I?* 3S ifí 5 S " r* v Z S o I 3 3 S í p- ~ fólii II - r - *■< <"> - » »ja 3s llE . 2 3i fe CD ■— — CD Ä C/1 O — Ľ3 03 72 co »t l VAUS, David de. 2002. Analyzing Social Science Data. London: Sage. 17 1 09 58 7402 58 Analysing social science data than a set level (e.g. 5 per cent or I per cent) the probability of sampling error is usually considered too low to worry about so the null hypothesis (o( no relationship in the population) is rejecled. The rejeclion of a null hypothesis lends support to ihe substantive hypothesis — that a difference between groups, a correlation or some other pattern at least as substantial as found in the sample, is likely to exist in the population. Should One-tailed or Two-tailed Significance Tests be Used? Tests of signinoncexome in two flavours: one-tailed and two-tailed. Since many statistics packages provide (lie option it is important to know what the difference is. '. i Choosing between one- and two-tailed tests depends on the nature of the substantive hypothesis being tested. We can distingidsh between directional and non-directional hypotheses. Hon-Directional Hypotheses Non-directional hypotheses predict that die sample value will simply be statistically different from a particular value. For example, we might anticipate that men and women will have different levels of marital satisfaction without predicting in which direction the difference will be - whether men or women will be more satisfied. We might predict that age and prejudice will be correlated without knowing whether older people will be less or more prejudiced than younger people. When your substantive hypothesis is non-directional, use a two-tailed test. Directional Hypotheses A directional hypothesis predicts an effect (correlation, difference between means of groups, etc.) end stales the nature of the difference - that the correlation will be positive, that group A will have a higher mean than group B (or whatever the predicted direction is). One-tailed tests are used for testing directional hypotheses. Notes 1 The null hypothesis does iwl have to be that Ht = 0. As G>hen (ISM) points out, we could begin with a proposition derived from a theory w previous research thai spořiti« a corsela-Uon of ii particular size and diiscllon. In (he Poppertan Uadifion of trying