Module 7 - MANOVA & MANCOVA Describe factorial designs and the extension of GLM to the multivariate case, limitations and examination of the tenability of the assumptions. Practice with data sets using PASWSPSS to run MANOVA and MANCOVA procedures. Continue practicing with data sets Chapter 6 and 7 of Cronk Pp 99 Practice Exercise: Using Practice Data Set 1 in Appendix B, determine if younger participants (< 26) have significantly lower mathematics scores than older participants. (Note: You will need to create a new ordinal variable representing each age group. You have conducted a study in which you collected data from 23 participants. You asked each subject to indicate his or her sex (SEX), age (AGE), and marital status (MARITAL). You gave each subject a test to measure mathematics skills (SKILL), where the higher scores indicated a higher skill level. The data are presented below. Note that you will have to code the variables SEX and MARITAL and also indicate that they are measured on a nominalscale: SEX M F F M M F F F M M F M F M M F F M F M M M F AGE 23 35 40 19 28 35 20 29 29 40 24 23 18 21 50 25 20 24 37 42 35 23 40 MARITAL Single Married Divorced Single Married Divorced Single Single Married Married Single Single Single Single Divorced Single Single Single Married Married Married Single Divorced SKILL 34 40 38 20 30 40 38 47 26 24 45 37 44 38 32 29 38 19 29 42 59 45 20 1 Pp 102 Practice Exercise Use the RACE.sav data file to determine whether or not the outcome of short- distance races is different from that of medium-distance races. Phrase your results. Section 7.5 Kruskal-Wallis H Test Description The Kruskal-Wallis H test is the nonparametric equivalent of the one-way ANOVA. It tests whether or not several independent samples come from the same population. Assumptions Because the test is nonparametric, there are very few assumptions. However, the test does assume an ordinal level of measurement for the dependent variable. The independent variable should be nominal or ordinal. PASW Data Format PASW requires one variable to represent the dependent variable and another to represent the levels of the independent variable. The dependent variable must be entered as a scale-type variable even if it is actually ordinal. Practice Exercise Use the RACE.sav data file to determine whether or not the outcome of short- distance races is different from that of medium-distance races. Phrase your results. Section 7.5 Kruskal-Wallis H Test Description The Kruskal-Wallis H test is the nonparametric equivalent of the one-way ANOVA. It tests whether or not several independent samples come from the same population. Assumptions Because the test is nonparametric, there are very few assumptions. However, the test does assume an ordinal level of measurement for the dependent variable. The independent variable should be nominal or ordinal. PASW Data Format PASW requires one variable to represent the dependent variable and another to represent the levels of the independent variable. The dependent variable must be entered as a scale-type variable even if it is actually ordinal. 2 Pp 110 3 Chapter 7: nonparametric inferential statistics - practice exercise: use the data in practice data set three in appendix B. If anxiety is measured on an ordinal scale, determine if anxiety levels change over time. Phrase your results. Practice Data Set 3 Participants who have phobias are given one of three treatments (CONDITION). Their anxiety level (1 to 10) is measured at three intervalsbefore treatment (ANXPRE), one hour after treatment (ANX1HR), and again four hours after treatment (ANX4HR). Note that you will have to code the variable CONDITION and indicate that they are measured on a nominal scale. ' .Type } Width I Decimals j ; Missing J[ Columns 1 anxpie Numeric 8 2 None None 8 3 Right f Scele \\ Input 2 anxlhr Numeric 8 2 None None 8 9 Right # Scale \\ Input 3 anx4hr Numeric 8 2 None None 8 = Right f Scale \\ Input 4 condit Numeric 8 2 |1 00. Place None 8 9 Right $s Nominal \\ Input Name Label j Values *'9" Measure Role Data sets chapters 4 and 9 of Mertler and Vannatta; these exercises pp 91 & 92; pp 261 &262 have been completed and posted. 4