Multiple Comparisons

Question 1

What is multiple comparison?

Answer 1

An easier way to determine where the difference lies b/w multiple groups.

Question 2

What does M.C. Control are?

Answer 2

Inflated p (of type erroe).

Question 3

Error rates?

Answer 3

The probability of type I error in specific analysis

Question 4

Error rate per comparison.

Answer 4

P. [ of type I error in a particular comparison]

Question 5

Pairwise comparisons.

Answer 5

• Diff btw 2 pairs of scores or means
  Max no. Of parrwise comp. K( k-1 ) /2

Question 6

Complex multiple comparisons!

Answer 6

Combined pairs of means are compared
I take an average of 2 groups& compare to
control.d

Question 7

Why is t-test problematic fork >2 design?

Answer 7

→ inflated TypeErrors as we set the error for each comparison we do b/w2 means
→separate error term. For each comparison could result in a diff t every time

Question 8

What are the 4 conceptual modifications applied to t-test to make it work for multiple comparisons?

Answer 8

① each mean is weighted based on its imp or role in analysis
② using MSe instead of pooled variance b/c pooled variance is an overestimate of population variance when k> 2
③ weights the no. Of groups in design.; includes no. Of groups in analysis & adjusts for unequal n’s
④ uses DF for for MSe team for critical value.

Question 9

What is a planned comparison? What is its advantage?

Answer 9

① done a-priori
② justifies our experiment ae why we are doing it
③ reflects the season for our study & why these levels of factor
④ Doesn’t require significant F
⑥ don’t have to do all possible comparisons.
→ pairwise or complex complex

Question 10

Why hy is planned comparison powerful?
.

Answer 10

→ more likely to reject null.
→ p alpha less stringent bid we are not doing au possible comparisons only selected few.

Question 11

Planned comparison - t-tests

Answer 11

→ if no. Of comparisons >l then set alpha PC
→ for every comparison will have a diff pooled variance.
→ loss of credibility
→ païnoise comparison >2 then use JASP or r
→ for both a prior or post-hoc

Question 12

What is the issue with t-test?

Answer 12

→ diff standard error for each comparison leading to diff values. For the t-obs

Question 13

What are the features of Dunn’s test?

Answer 13

① sig f not needed
② p-alpha (ew) is based on #comparisons
③ credible & I conservative
④ pooled error term → MS e
⑤ critical value based on no. Of comparisons made & not all possible

Question 14

Dunn’s special t-distribution is based on……..

Answer 14

Bonferroni’s t
→ the probability of one or more events cannot exceed the sum of individual probabilities
→ error rate can never exceed alpha
→p alpha per comparison = p- alpha ew/ #comparisons

Question 15

Steps for Dunn’s t-test

Answer 15

① apply test ratio to find t- durns value
②diff b/w of groups
③ compare to i get a decision.

Question 16

Why do we do post-hoc comparisons?

Answer 16

→ study is novel/ not well researched
→ no basis for prediction
→ exploratory analysis
→ unexpected outcomes
→ found nothing in planned comparisons
→ requires a sig f-obs
→ adjust p-alpha ACC. To all comparisons (total)
→ less powerful alternative I hard to reject null b/c individual p-alpha values are small

Question 17

Fishers LSD test

Answer 17

→ least sig differ min diff b/w - means required for significance
→ sig f-obs
→ k= ‘ 3 design is best for this test
→ homo. Of variance can’t be violated.

Question 18

Describe Tukey’s HSD

Answer 18

→ requires sig f-obs
→ best for k > 3 design
→ set p-alpha (ew)
→ q-table (a-10)
→ q-crit involves, alpha, k, DFe

Question 19

What are the constraints applied to dunnets?

Answer 19

① can only do k-1 comparisons
② have to compare to only control group.
③ used when homogeneity of variance is violated

Question 20

What all to include in the formal report for multi comparisons

Answer 20

① Levene’s
②mean & SD of each group
③ anova statistics F, MS e, eta sq.,p, cohen’s F
④ planned comparison
⑤ post-hoc - but only condition not reported in ④. If reporting both P.C.& post-hoc

Question 21

What are the steps for post-hoc probes - computation?

Answer 21

①identify degrees of freedom, alpha,MSe,
② identify K
③ critical value from diff table depending on the test
④ compute test, ratio
⑤ diff of means bow possible means
⑥ diff of means
⑦ compare means to critical value