Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
58 Cards in this Set
- Front
- Back
example of an aposteriori test
|
Scheffe's test
|
|
tests subsequent to ANOVA allow more _______ tests
|
precise
|
|
sample null hypotheses for tests subsequent to ANOVA
|
m1+m2/m3 = 0
m1 - m2 - m3 = 0 |
|
How many comparisons are allowed in orthogonal comparisons?
|
as many comparisons as the degrees of freedom within; g-1 comparisons
|
|
What is the most powerful test subsequent to ANOVA?
|
orthogonal comparisons!
|
|
t/f Scheffe's test is the least powerful test subsequent to ANOVA.
|
true
|
|
DF in what comparison are always 1?
|
BONFERRONI, DOODOO HEAD!!
|
|
is bonferroni mutually orthogonal?
|
negative, soul sistah.
|
|
t/f: in orthogonal comparisons, you can conduct more comparisons that DF.
|
false, in bonferroni tests you can make more comparisons than DF.
|
|
what is the familywide error rate?
|
the probability that a family of conclusions will have AT LEAST one type I error.
|
|
P(no type I error) =
|
1 - alpha
(alpha = .05 unless otherwise specified) |
|
P(type I error) =
|
alpha
|
|
Bonferroni correction:
|
alpha/K
(k is the number of intended comparisons) |
|
are Scheffe comparisons orthogonal?
|
nope. (Ea = 0, but Eab does not = 0)
|
|
Which test has unlimited comparisons?
|
Scheffe's!
|
|
Scheffe's familywise error rate is
|
= .05
|
|
the two range tests are ___ and ___; they are aposteriori/apriori
|
newman-keul and Tukey test
APRIORI! |
|
how do the two range tests differ?
|
Tukey test g = number of groups in experiment; Newman-Keul g = number of groups in specific comparison; changes critcal q value
|
|
why are tukey and newman-keul called range tests?
|
they order means from low to high
|
|
familywise error rate in tukey
|
= alpha, usually .05
|
|
range tests are based on what kind of distributions?
|
t distributions
|
|
Why did Newman-Keul adapt the Tukey test?
|
said means with greater differences were tested with more power than those with smaller differences
|
|
which test is more liberal, tukey or nk?
|
nk tests are more liberal, meaning there is a slightly higher chance of type I error
|
|
btw subjects independent variables
|
each subj receives only one level of an IV (positive OR neg mood, male OR female)
|
|
wi subjects indep variables
|
each subject receives all levels of an IV (pos, neg, neutral mood treatments, interacts with m and f)
|
|
simple effect:
|
measure of an independent variable at a single level of a second independent variable
|
|
two way interaction:
|
extent to which simple effects are different in direction and/or size
|
|
confound:
|
varies systematically along levels of an independent variable
|
|
repeated measures: within subjects IV
|
participants receive each level of an IV
|
|
repeated measures: fixed effect IV
|
experimenter controls which levels will be used (decides to observe happiness and sadness vs picking two moods out of a hat)
|
|
repeated measures: random effect IV
|
experimenter allows IV to vary by chance (picks two moods to compare from a hat randomly)
|
|
are subjects a dependent or independent variable in repeated measure designs?
|
independent variable
|
|
why is subjects' variability not considered error in repeated measures designs?
|
in repeated measures designs the subjects are considered an independent variable, so the amount they are measured is the fixed variable and subjects and the random effect iv.
|
|
which is the error term in a repeated measures (within subjects) design?
|
sxa interaction, because we can't explain why the scores are the way they are; the effect of A across S is not systematic
|
|
repeated measures assumptions:
|
no carry-over effects!
|
|
how do you account for specific (known) carryover effects?
|
return subjects to baseline (bring eyes back to regular light exposure)
|
|
how do you account for general carryover effects (effects due to repeated measure)?
|
give subjects breaks or practice trials
|
|
t/f specific carryover effects are due to repeated measure
|
false; general carryover effects due to repeated measure
|
|
violations of what assumption increase type one error?
|
the assumption that there will be equal measure to measure correlation in repeated measures designs
|
|
equal measure to measure correlation:
|
correlation between any pair of trials is approximately the same
|
|
solutions to violations of equal measure to measure correlation
|
1. MANOVA: treat each level as though its IVs are separate and dependent vars
2. multiply box's epsilon by any within DF, making the test more conservative |
|
SS in this test do not substitute for MS
|
scheffe! only substitutes in orthog and bonferroni
|
|
whats the difference between orthog comps and bonferroni?
|
bonferroni acounts for a familywise error rate (alpha/k) and is more conservative as a result
|
|
DF = g - 1 for these tests:
|
orthogonal, scheffe
|
|
df = 1 for this test
|
bonferroni
|
|
2 way btw grps ANOVA
DF: |
rows, columns, rxc interaction
Total: N-1 Btw: rc - 1 row: r - 1 col: c - 1 rxc: (r-1)(c-1) within N - rc |
|
within grps ANOVA:
DF: |
A, SxA
total: N-1 btw S-1 wi: N - s A: a - 1 AxS: (s-1)(a-1) |
|
in wi grps ANOVA, what do these letters stand for?
a: S: A: AxS |
a: number of scores in each S
S: row total A: column total AxS: error within |
|
how do you look up critical F in SxA?
|
DF for A, DF for SxA
|
|
which f ratio do we care about in SxA?
|
A over SxA
|
|
why is unequal n a problem?
|
1. confound creates a conceptual problem
2. sums of sqaures of the cells do not equal the sums of squares of the main effects and sums of squares of interactions 3. difficulty in deriving orthogonal coefficients |
|
box's epsilon =
|
1/ t -1
|
|
assumptions of repeated measures ANOVA
|
1. there are no carryover effects between trials
2. there is equal measure to measure correlation (correlation btw any pair of trials is approximately equal) |
|
simple effect:
|
effect of one IV at a single level of another IV (i.e. does m at positive mood affect score?)
|
|
main effect:
|
effect of one IV ignoring the others (gender ignoring mood)
|
|
2 way interaction:
|
extent to which simple effects are different in direction and/or magnitude (ex. does mood affect score for f differently than m?)
|
|
between subjects independent variables:
|
each subject receives only one level of the IV (pos OR neg)
|
|
within subjs IV:
|
each subject receives all levels of an IV (3 levels of abc)
|