For example, when we conduct a 2GROUP(A, B)x 2varience(C, D)mixed repeated ANOVA, we found that condition C is significantly larger than condition D in the two groups. However, the η2 of the F -TEST in GROUP A (F = 5.26，p = 0.026，η2p = 0.019) is smaller than GROUP B(F = 23.01，p < 0.001，η2p = 0.197). Can we say that the effect of variance on B is middle size and is greater than that of A?
There isn't enough information because we don't know the means or which factor repeated measures. However, it is worth noting that η2p can't be used to compare the size of effects between different types of factor because the denominator in the calculation is different. Its almost like comparing meters/second and meters/minute. Specifically the n2p for the repeated measures factor excludes individual differences, but it includes it for the independent measures factor. So the mean difference in one factor could be bigger than that in the other and yet the n2p value be smaller.
You can change this by reporting generalised eta squared instead of partial eta-squared.
That doesn't mean n2p is completely useless. It can be used in power calculations as it accounts for the superior precision you typically get in repeated measures designs. However, this assumes that you are using a similar design in future. If you switched design from a mixed to independent measures one the n2p value would still be misleading.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA