Sie sind auf Seite 1von 5

1

One-way ANOVA Multiple comparison tests (DR SEE KIN HAI)


1. If you want to compare more than 2 groups of means of an independent variable use 1-way ANOVA 2. Basic procedure: To get 2 different estimates (between-groups variance a measure of the effect of the independent variable combined with error variance; within-groups variance a measure of error variance; variability of scores around the mean of a variable) of population variance from the data and calculate a statistics from the ratio of the 2 estimates F-ratio = btw gp var / within gp var. 3. If F-value is significant = population means are different so null hypothesis (all means are equal) is rejected. To locate where the sig differences lie use post-hoc analysis. 4. Post-hoc analysis go through data hunting for whatever sig might be there preventing Type 1 errors (Ho is true but rejected). Scheffe test allows you to perform every possible comparison but is tough on rejecting Ho. Tukeys Honesty Sig Diff (HSD) test is more lenient but restricted to types of comparison. Example: If you want to compare the effects of using 3 different teaching methods on 3 different groups of students academic scores as shown in the table below, use a 1-way ANOVA multiple comparison tests. Group 1 (Teaching Method 1), Group 2 (Teaching Method 2), Group 3 (control) 11 8 3 12 9 6 10 7 3

2 How to run 1-Way ANOVA Multiple Comparison Tests 1. Select [Analyze] then [Compare Means] and [One-Way ANOVA..] to open the dialogue box below.

2. Move [Score] into the [Dependent List] and [Groups] into the Factor] box and select [Post Hoc..] to open the sub-dialogue box.

3. Select [Scheffe], [Tukey] and [Duncan] tests then [Continue] and [OK] Interpreting the output

The table shows the results for ANOVA. F ratio = 22.200 for the between groups effect (i.e. the effects of different Teaching Method) which has an exact significance level of 0.002. Thus, the between groups effect is significant implying the means for the 3 groups differ
Here, Group 2 and Group 3 are in the same subset group, so no sig different btw them but both Groups are sig diff with Group 1

e.g. using Tukey HSD test, Teaching Method 1 is not significantly different from Method 2 (p = 0.066) but is significantly different from the Control Group Mean (p = 0.001)

Here, Control Group is the only group in this first subset. Consequently it is significantly different from the other 2 group means

Tukey HSD and Scheffe Tests show similar results Group 2 and Group 1 are in the same subset i.e. they are not significantly different from each other but both Groups are significantly different from the Control Group

Duncan test shows that the Control Group is sig different from Group 2 mean and Group 2 is sig different from Group 3 mean

4 Reporting the output A one-way unrelated analysis of variance showed an overall significant effect for the different teaching methods (F ratio = 22.20, p = 0.002). Scheffes range test found Group 1 did not differ significantly from Group 2 (p = 0.077) but differed significantly from Control Group (p = 0.002).

COURSEWORK You, as a teacher wanted to investigate the effectiveness of using 3 different teaching methods (Method 1,2,and 3) on students academic achievement (on any subject). And also you wanted to find out which Teaching Method is the most favorable to the students. A quasi-experimental research design has been chosen for this study. By using the intact classes, you have divided the students into three Groups (A, B, C). You have administered a pretest to all the 3 groups. Group A were taught using teaching method 1, Group B with method 2 and Group C with method 3. After three weeks, all the 3 groups sat for the posttest and the scores for the posttest are shown in the Table below. Table below shows the posttest scores results for the three Groups Group A Group B Student Method Scores Student Method Scores 1 1 60 1 2 70 2 1 65 2 2 72 3 1 55 3 2 85 4 1 50 4 2 84 5 1 58 5 2 82 6 1 62 6 2 78 7 1 68 7 2 88 8 1 70 8 2 74 9 1 52 9 2 80 10 1 62 10 2 76

Group C Student 1 2 3 4 5 6 7 8 9 10

Method 3 3 3 3 3 3 3 3 3 3

Scores 80 82 80 90 92 98 95 90 95 90

Das könnte Ihnen auch gefallen