120 likes | 145 Views
Statistics in SPSS Lecture 9. Petr Soukup, Charles University in Prague. ALTERNATIVES TO T-TESTS (NON-PARAMETRIC TESTS). Alternatives to T-tests. Alternatives in SPSS. Analyze-Nonparametric tests-One sample Analyze-Nonparametric tests-Independent samples Analyze-Nonparametric tests-Paired
E N D
Statistics in SPSSLecture 9 Petr Soukup, Charles University in Prague
Alternatives in SPSS Analyze-Nonparametric tests-One sample Analyze-Nonparametric tests-Independent samples Analyze-Nonparametric tests-Paired Example: alternative to ANOVA: Analyze-Nonparametric tests-Independent samples-Compare Medians (Kruskal-Wallis with multiple comparisons)
Contingency table (CT) Combine at least two (nominal or ordinal) variables in one table Main goal: to analyze relationship between variables Descriptive stats in CT: observed counts and percentages (row, column and total) Analyze-Descriptive statistics-Crosstabs (Cells)
Test for CT Test of independence: Chi-square H0: variables are independent (in the whole population) H1: variables are dependent (in the whole population) Analyze-Descriptive statistics-Crosstabs (Statistics) Computation: comparison of observed and expected (by H0) counts Note: for correct computation 80 % of exp. Counts>5 and no zeros
Contingency coefficient (CC) Effect size for CT Measurement of relationship between variables Usually computed from chi-square Values between 0 (no relationship) and 1 (perfect relationship) Mostly used coefficient: Cramer’s V Analyze-Descriptive statistics-Crosstabs (Statistics)
Detailed analysis in CC For proven relationship (reject H0) Sign scheme for adusted standardized residuals Once again rule of values -2 and +2 Signs instead of residuals and it’s meaning Analyze-Descriptive statistics-Crosstabs (Statistics) and MS Excel tool for sign scheme
HW9 Try to analyze relationship between two nominal or ordinal variables by contingency tables. Interpret results.