Exam 2: Multiple Comparison Tests (Lec 6)

Multiple comparison tests

Companion statistics that help order the means and give comparisons between each two pair of means

This nonlinear trend involves 2 changes in direction

Cubic trend

This nonlinear trend involves 3rd turn

Quartic trend

Familywise error rate

Represents the probability of making at least one Type I error in a set or "family" of statistical comparisions

Which of the following tests gives the greatest control over Type I error, but at the expense of power? a) Scheffe's b) Bonferroni c) Newman-Keuls d) Tukey's

a) Scheffe's

The Bonferroni t-Test splits alpha evenly among the set of planned contrasts and this is known as Bonferroni's _____, and is used to protect against Type __ error.

adjustment, I

The Bonferroni's adjustment is the process of adjusting ___, and is used to protect against Type __ error.

alpha, I

Trend analysis is called ____ when all data rise or fall at a constant rate as the value of the independent variable increases

linear

When analyzing multifactor designs, the multiple comparison procedures can be used to compare means for both ___ effects and ____ effects.

main, interaction

Analysis of significant main effects involves comparison of differences between ____ means.

marginal

When a single experiment involves at least one independent factor and one repeated factor, the design is called a _____ design.

mixed

Trend analysis is _____ when it demonstrates bends or changes in direction

nonlinear

Newman-Keuls method uses a _____ error rate where alpha specifies Type __ error rate for each pairwise contract.

per comparison, I

True or False: One-tailed test can be performed for planned comparisons even with weak evidence for a directional hypothesis

False; the evidence should be quite strong

True or False: The purpose of trend analysis is to describe the overall tendency in the data using the most number of trend components as possible

False; use the least

True or False: Like Tukey, Newman-Keuls method uses only one critical difference for all comparisons.

False; uses a larger critical difference as r increases

A conservative test will protect against Type __ error, requiring that means be ____ to establish significance.

I, far apart

This nonlinear trend demonstrates single turn upward or downward (creating a concave shape to the data)

Quadratic

Per comparison error rate

Represents the probability of making a Type I error for once comparison

True or False: ANOVA only tells us that there is a significant difference between at least 2 means amongst 3 or more groups

True

True or False: Beyond the highest and lowest means being different, ANOVA doesn't tell us which other groups are significantly different from each other

True

True or False: Generally, higher order nonlinear trends are more complex and often very difficult to characterize or intepret

True

True or False: In regards to the familywise error rate, as the number of means goes up, so does the probability of a Type I error.

True

True or False: Post hoc comparisons usually use two-tailed probabilites

True

True or False: Regarding Newman-Keuls method, as the number of comparisons increase, the chances of committing a Type I error are greater.

True

True or False: When comparisons are planned in advance and when they are relatively limited in number, a prior tests can be used.

True

The major difference between multiple comparisons tests lies in the degree of protection offered against ____.

Type I and Type II error

For all the types of tests, the general process is similar: First, the means are arranged in ____ order of size and ____ between each pair of means is obtained; Next, each pairwise comparison is test against a ____ _____ _____.

ascending, difference, minimum significant difference

Which of the following tests is the most powerful of the post hoc comparisons? a) Scheffe's b) Bonferroni c) Newman-Keuls d) Tukey's

c) Newman-Keuls

Which of the following tests, over the long run, will produce more Type I errors because it does not control for family wise error rate? a) Scheffe's b) Bonferroni c) Newman-Keuls d) Tukey's

c) Newman-Keuls

A more liberal test will find significant difference with means that are _____, thereby offering greater protection against Type __ error.

closer together, II

Trend analysis aims t find the most reasonable description of _____ data based on the number of "ups and downs".

continous

Which of the following correctly states how multiple comparison tests calculate Type I error rate? a) All will use comparison error rates b) All will use familywise error rate c) All will do t-tests d) Some will use per comparison rates and some will use familywise error rate

d)

Which of the following tests is often preferred because it offers both reasonable power and protection against Type I error? a) Scheffe's b) Bonferroni c) Newman-Keuls d) Tukey's

d) Tukey's

Calculation of MSD is based on ____ ____ ____, from the ANOVA and a critical value taken form a statistical table.

error mean square

Critical values are located according to degrees of freedom associated with the ____ ____, in the analysis of variance.

error term

Regarding Tukey's, alpha identifies the probability that one or more of the pairwise comparisons will be ____ declared significant and there is generous protection against Type __ error.

falsely, I

The Bonferroni t-Test uses a ____ error rate that is the sum of the per comparison significance levels.

familywise

The Scheffe comparision is based on the F-distribution and is CONSERVATIVE in that it adopts a _____ error rate that applies to all contrasts, giving strong protection against Type __ error but also makes it less _____ than the other tests.

familywise, I, powerful

Tukey's Honestly significant difference test uses a _____ error rate and is calculated using the _____ range statistic given the symbol __.

familywise, studentized, q

If the absolute difference between a pair of means is equal to or ___ than the minimum significant difference, then the contrast is considered SIGNIFICANT.

greater

As the number of sample means increases, the size of the difference between the largest and smallest means will ____, even when the null hypothesis is ____.

increase, true

Some multiple comparison tests are classified as ______ comparisons (a priori) where specific relationships planned ____ to data collection based on the research rationale.

planned, prior

Most multiple comparison tests are classified as ____ ____ because done after the analysis of variance is completed.

post hoc

Multiple comparison tests are considered _____ comparisons because only done after initial exploration of the outcomes by ANOVA, and most useful when a general alternative _____ has been proposed.

unplanned, hypothesis

In regards to the concept of repeated measure analysis, when all factors are repeated this is referred to as a repeated measure or _____ design.

within-subject