1/16
Looks like no tags are added yet.
What inferential procedure should follow a significant ANOVA F test?
the Scheffe method - is an ANOVA simultaneous test procedure and an associated simultaneous confidence interval method.
This is the appropriate follow-up approach for a post-hoc contrast analysis.
Describe the Scheffe method?
The scheffe method uses the ANOVA F test as a basis for a simultaneous test procedure, allowing for a test on all hypotheses implied by the homogeneity hypothesis
What is the decision rule associated with the Scheffe method?
Given
The decision rule fo the scheffe method is as follows
What are the steps associated with the Scheffe method?


How to calculate the confidence interval given a contrast?


Contrast simple vs. complex contrast?
A complex contrast is defined as a contrast for three or more groups,
Whereas a comparison includes data for only 2 groups only.
Compared to a simple contrast, a complex contrast has smaller - smaller Σc2 and SE, and hence greater precision for CI (i.e., narrower CI)
• more statistical power to detect a TE (for test of H0
ψ = 0).
What is the maximal contrast


What is the maximal contrast used for?
SHOULD NOT be used for the test of F test of homogeneity hypothesis – as the conclusions are generally not interpretable
HOWEVER,
You can use it to suggest interpretable (m,r) contrasts → if the maximal contrast is significant then the (m,r ) contrast with the largest SS, will be those with a coefficient.
E.g
Describe ANOVA output for contrast and note contrast between PSI max and ANOVA F.
PSI max contrasts’s significance cannot be derived from the ANOVA F contrasts as df is different for each case.
What is a redundancy?
A contrast inference is redundant if it is implied by the reference for two (or more) contrasts in the analysis .
E.g
What is a contradiction?
a contrast inference may be contradicted by the inference for two (or more) or other contrasts.
If we end up in situation, treat the significant outcome as type 1 errors, or if we should treat non-significant outcomes as type 2 errors.
The Scheffe wise error rate is able to control ALL types of type 1 errors. – therefore you can exclude contradictions.
What is the relationship between contrast complexity (confidence intervals) and precision (in comparison a simple confidence interval)?
The more complex a contrast is, the more precise that contrast will be in reference to a standard simple confidence interval.
This is because for a complex contrast, the Σc2 is generally smaller than simple contrasts (meaning that SE is smaller), and therefore more precise to the actual range of the true population parameter
Therefore it was more power to detect TE (when null = 0 )
What is the relationship between the ANOVA and the Scheffe procedure?
Can be used together to control the FWER in a post-hoc analysis.
What are the properties of the F STP contrast?
It controls the FWER error rate for any contrast analysis
It is a coherent analysis – implied hypothesis cannot be rejected
The analysis is consonant – if the overall H(0) can be rejected then at least one of the follow up contrasts can be as well. * may or may not be interpretable
Conversely, if the overall ANOVA does not reject the null hypothesis, the Scheffé test will not find any significant contrasts.
The order of the test contributes nothing
STP inference are compatible the simultaneous confidence produced by the F-based Scheffe SCI procedure
It controls the familywise non-coverage error rate at Ci on all contrasts
It is completely compatible for the ANOVA F test.
Is the Scheffe test conservative or not?
is conservative for analyses restricted to only mean difference {m, r} contrasts, or to only comparisons (or to only a subset of each). In this context, conservative implies that the actual FWER is much smaller than nominal α.
What is the main advantage of the Scheffe post-hoc analysis is its flexibility?
no restriction on number and nature of contrasts
can define contrasts based on the pattern of group means and/or coefficients of maximal contrast.
HOWEVER, this means that this type of post-hoc analysis is conservative when restricted to only means difference or only comparisons – the actual FWER is much smaller than nominal alpha