Lecture 8 - ANOVA 2

Question 1

A follow up test uses ____ ____

Answer 1

post hoc

Question 2

Suppose the Ho is rejected, what might we want to know ?

Answer 2

We might want to explore which pairs were responsible for the rejection.

Question 3

How will we answer the question: Which pairs were responsible for the rejection ?

Answer 3

We will study 3 “adjustments” to the 2 sample t-test (assuming equal variances to answer this question).

Ho: Mu(i) = Mu(j)
Ha: Mu(i) does not equal Mu(j)

Question 4

LSD

Answer 4

least significant difference

Question 5

N

Answer 5

total number of observations

Question 6

k

Answer 6

the number of groups

Question 7

look at formulas on slide 7

Answer 7

okay

Question 8

What is the Bonferroni Correction ?

Answer 8

We modify the level of significance by dividing by the number of possible pair wise comparisons, m.

m = (k/2) = k(k-1)/2

*If m = 10, alpha = 0.005, our chance of making an error is 0.05

Question 9

HSD

Answer 9

honestly significant difference

Question 10

The q tables take into account ______

Answer 10

bonferroni

The q table just simplifies the bonferroni approach.

Question 11

How do you do turkey multiple comparison test ?

Answer 11

You subtract the means and then see how they compare to HSD

You determine HSD by doing q (square root of MSW/n

Question 12

What are Posterior Unplanned Comparisons?

Answer 12

• Tukey’s HSD Test
• LSD Procedure (Fishcer’s)
• Newman-Keuls
• Duncan’s Test
• Scheffe’s

Unplanned: This one you’re comparing every possible comparison

Question 13

What are A Priori Planned Comparisons?

Answer 13

• Multiple t tests
• Dunn’s (Bonferroni)
• Linear Contrasts

Planned: You have some pretty specific questions. Don’t care about comparing every possible comparison.

Question 14

Linear Contrasts:

All the weightings have to add up to ___

Answer 14

zero

Question 15

Linear Contrasts:

What do you weight the ones that you don’t care about?

Answer 15

zero

Question 16

Linear Contrasts:

When asking questions and you have the appropriate weightings, what do you calculate and how do you interpret it?

Answer 16

Do (weighting x mean)

add them all up and square them

then times by n

divide by the sum of weighting^2

Question 17

Linear Contrasts:

Once you have the number to the previous calculation, how do you get the F value ?

Answer 17

Divide by MSW

*Then compare the calculated F value to the critical F value and answer the original questions that were asked

Question 18

What is key when determining linear correlation between data ?

Answer 18

Need to do equally spaced weightings to determine if there’s a linear correlation between the data.

Question 19

How many questions are you allowed to ask?

Answer 19

You’re allowed to ask 1 less than the number of groups.

So for the smoking question, we had 4 groups, so we can ask 3 questions.

Question 20

The questions need to be _______ of each other.

Answer 20

independent