Topic 7: ANOVA Flashcards
What research design uses one-way ANOVAs?
K > 2 (over two sample groups)
What is the null hypothesis being tested by ANOVA?
Mean-1 = Mean-2 = Mean-3
What is wrong with conducting multiple t-tests?
It increases the likelihood of Type 1 error (rejecting the null hypothesis when it is true)
What are the assumptions of an ANOVA test (3)?
Scores in each sample - come from pop where DV is normally distributed
Homogeneity of population variances
Random sampling and random assignment of data
What is MSerror, and is it affected by the accuracy of the null hypothesis?
Within-groups estimate of the population variance
Not affected by the accuracy of the null hypothesis, as it focuses on the variation inside each population
What is MSgroups, and is it affected by the accuracy of the null hypothesis?
Between-groups estimate of the population variance
If it’s true, the samples will have no difference, as they’re drawn from identical populations
If it’s false, it’s influenced by variation within the population AND variations among the population means
What are the characteristics of the distribution when you have more than 2 groups?
Random sampling model Independent samples research design with 1 factors, k = 3 independent groups Score data Statistic: group means Pop parameters: not known Sample size: nj = X (N=Y)
What are the variances that the f-statistic is comprised of?
MSgroups (numerator) and MSerror (denominator)
What are the characteristics and shape of the sampling distribution of the f-statistic?
Distribution of the f-statistics for df of [dfgroups] and [dferror]
Shape: asymmetric, positively skewed
X-axis: significant outcomes at positive end of tail
- no upper limit of values
CT: if equal ns, median = 1.00
Family/series of curves: 2 calculations from df
What are the advantages of a k >2 design?
Advantages:
- Can analyze >2 groups while controlling for TYpe 1 errors
- Better interpretation of the impact of the IV on the DV
- More powerful: less unexplained variability, pooled variability
How are the degrees of freedom of between-groups found?
k -1
How are the degrees of freedom of within-groups found?
N - k
How are the degrees of freedom of within-groups found?
N - k
What variables does Fcrit vary with (3)?
Fcrit varies a function of alpha, dfgroups and df error
What are the disadvantages of a k>2 design?
Comparing several condition means - don’t know where the difference is
Need to still apply multiple comparisons
