two-way between subjects anova also called: two-way randomized anova also called: two-way randomized...
Post on 19-Dec-2015
214 views
TRANSCRIPT
TWO-WAY BETWEEN SUBJECTS ANOVA
Also called: Two-Way Randomized ANOVA
Purpose: Measure main effects and interaction of two independent variables
Design: factorial Assumptions: same as one-way BS
ANOVA
Dividing the Variance Total = A + B + AxB + Within
Groups A: differences between levels of A B: differences between levels of B AxB: other between group
differences Within Groups: differences within
groups
A Variance affected by:– effect of Factor A (systematic)– individual differences (non-
systematic)– measurement error (non-systematic)
B Variance affected by:– effect of Factor B (systematic)– individual differences (non-
systematic)– measurement error (non-systematic)
AxB Variance affected by:– AxB interaction (systematic)– individual differences (non-
systematic)– measurement error (non-systematic)
Within Groups variance affected by:– individual differences (non-
systematic)– measurement error (non-systematic)
Comparing the Variance
FA MSA
MS Within
FB MSB
MS Within
FAxB MSAx B
MS Within
ANOVA Summary Table
Source SS df MS F p
Factor AFactor BAxBWithinTotal
EXAMPLE: An oral or written spelling test was given in one of three noise levels. Determine whether there were significant effects of test type, noise level, and the interaction of test type with noise level. (See data on next page)
Computation of Two-Way BS ANOVA
Noise LevelNo Low High
oral15 15 12
Test 17 19 10
written 18 14 1014 12 12
Noise LevelNo Low High
oral 16 17 11 14.67Test
written 16 13 11 13.33
16 15 11
Overall mean = 14
Means
Source SS df MS F pTest TypeNoise LevelTest x NoiseWithin Total
ANOVA Summary Table
STEP 1: SS Between = S(xc-x)2
condition mean
xc xc-x (xc-x)2
16 2 416 2 417 3 917 3 911 -3 911 -3 916 2 416 2 413 -1 113 -1 111 -3 9 11 -3 9 SS Between = 72
STEP 2: SS A = S(xa-x)2 (A is Test Type) mean for level of A
xa xa-x (xa-x)2
14.67 .67 .4514.67 .67 .4514.67 .67 .4514.67 .67 .4514.67 .67 .4514.67 .67 .4513.33 -.67 .4513.33 -.67 .4513.33 -.67 .4513.33 -.67 .4513.33 -.67 .45 13.33 -.67 .45 SS Test Type = 5.40
STEP 3: SS B = S(xb-x)2 (B is Noise Level) mean for level of B
xb xb-x (xb-x)2
16 2 416 2 416 2 416 2 415 1 115 1 115 1 115 1 111 -3 911 -3 911 -3 9 11 -3 9 SS Noise Level = 56
STEP 4: SS AxB = SS Between - SS A - SS B
SS Test Type x Noise Level =72 - 5.40 - 56 = 10.60
STEP 5: SS Within = S(x- xc)2
x x-xc (x-xc)2
15 -1 117 1 118 2 414 -2 415 -2 419 2 414 1 112 -1 112 1 110 -1 110 -1 1 12 1 1 SS Within = 24
Source SS df MS F pTest Type 5.40Noise Level 56.00Test x Noise 10.60Within 24.00Total 96.00
ANOVA Summary Table
STEP 6: Calculate degrees of freedom.df A = a-1 a= # levels of A
df Test Type = 2-1 = 1
df B = b-1 b= # levels of Bdf Noise Level = 3-1 = 2
df AxB = (a-1)(b-1) df Test x Noise = (1)(2) = 2
df Within = (a)(b)(n-1) n = # per group df Within = (2)(3)(1) = 6
Source SS df MS F pTest Type 5.40 1Noise Level 56.00 2Test x Noise 10.60 2Within 24.00 6Total 96.00 11
ANOVA Summary Table
STEP 7: Calculate Mean Squares.
MS Test Type = 5.40/1 = 5.40MS Noise Level = 56/2 = 28.00MS Test x Noise = 10.60/2 = 5.30MS Within = 24/6 = 4.00
Source SS df MS F pTest Type 5.40 1 5.40Noise Level 56.00 2 28.00Test x Noise 10.60 2 5.30Within 24.00 6 4.00Total 96.00 11
ANOVA Summary Table
STEP 8: Calculate F-ratios.F(Test Type) = 5.40 /4.00 = 1.35F(Noise Level) = 28/4.00 = 7.00F(Test x Noise) = 5.30/4.00 = 1.32
STEP 9: Look up critical values of F.Test Type F-crit (1,6) = 5.99Noise Level F-crit (2,6) = 5.14Test x Noise F-crit (2,6) = 5.14
STEP 10: Compare F to F-crit. If F is equal to or greater than F-crit, reject the Null Hypothesis.
Test Type 1.35 < 5.99 Not sig.Noise Level 7.00 > 5.14 Sig.Test x Noise 1.32 < 5.14 Not sig.
Source SS df MS F pTest Type 5.40 1 5.40 1.35 >.05Noise Level 56.00 2 28.00 7.00 <.05Test x Noise 10.60 2 5.30 1.32 >.05Within 24.00 6 4.00Total 96.00 11
ANOVA Summary Table
APA Format Sentence
A Two-Way Between Subjects ANOVA showed a significant main effect of Noise Level, F (2,6) = 7.00, p < .05, a nonsignificant main effect of Test Type, F (1,6) = 5.40, p > .05, and a nonsignificant interaction, F (2,6) = 1.32, p > .05.
Computing Effect Size
Compute 2 for each effect
2 =SS Effect
SS Total
2 for A =SS A
SS Total
2 for B =SS B
SS Total
2 for AxB =SS AxB
SS Total
2 for Test = 5.40
96.00 = .06
2 for Noise =56.00
96.00 = .58
2 for Test x Noise =10.60
96.00 = .11