ANOVA Flashcards
ANOVA
assessing stat evidence btween 3 or more group means
t tests
assessing stat evidence between 2 or less group means
CR-p
one-way anova. 1 IV- Dv can be multiple
CRF-pq
two-way anova. 2 IV’s. determining interaction
fixed effects model
all tx levels included
random effects model
only p level of models included
Type 1 error
null hypothesis is true and rejected
type 2 error
a false hypothesis is not rejected
Increased type 1 error
decreased type 2 error and power
increased type 2 error
decreased type 1 error, increased power
f stat
variability between groups/variability w/in groups
mean 1 = mean 2
null
mean 1 does not equal mean 2
alternate hypothesis
F stat apa form
F(df1,df2)=fstat, P< 0.05
between group effects
effects between tx groups have no commonality
trend analysis
ordering (ordinal)
regression
continious IV and DV
pairwise
includes all group means
nonpairwise
2 or more, but not all group means
orthogonal
product of all coefficients = 0
RM
one within. subject recieves all tx levels
subject by treatment
order of tx levels is random
subject by trial
same order of tx levels
effect size
standarized diff of means across groups
CRF-pq
subjects in 1 level of 1st factor & one level of another
IV -surface (grass, clay) IV2-stroke (backhand, forehand)
two-between
No interaction in two-between
Main effects
fixed effects hyptoth
means across all levels of A are =, disregarding B
Significant interaction
Simple main effects
simple main effects
looks at one factor with levels of another Use pairwise comarisons
FMF
two within. each person will have all levels of tx’s
SPF
1 btwn 1 within, recieve one level of between and all levels of within
Post Hoc SPF
only for between effects
ANCOVA
analyze effet of IV on DV when controlling for a continous variable
Correlation increases
power
Which variables are continous in ANCOVA?
DV & cov
covariable is reasonably correlated with
DV
Effects of training on problem solving when controlling for thinking skills?
covariate
ANCOVA pros
decrease error due to covariate, decrease variability that cannot be controlled
Bonferroni confidence itnerval adjustment
controls for type 1 error
CRF Assumptions
no outliers, homogenity of variance, indepence, normality
outlier z-score
remove greater than 2.5
levene’s
homogeneity of variance
violation of homo
welch’s
shapiro wilkes
normality
kruskall wallis
correction for violation of normality & homo
Kruskall-wallis APA stat
X^2(DF)= stat, p <0.05
Post Hoc Tests with no violatino
LSD, Tukey, Tukey-Kramer, Gabriel, hochbergs, sidak, bonferroni, scheffee, dunnett
Post Hoc Tests with violation
Games-Howell, Dunnet’s C, Dunnet’s T3
Greater power, greater type 1 error
liberal post hoc tests
RM & RMF assumptions
independece, normal distribution, homogeneity of variance, sphericity
sphericity
homo. of variance w/in groups
Machley’s
test of sphericity
P > 0.05 assumptions are good
levene’s, shapiro wilkes, mauchley’s
WHen to use Greenhouse Geiser
if GG is < 0.75
Hyn Feldt
if GG is > 0.75
Contrasts
polynomial, difference, simple, deviation, helmert, repeated
CRF-pq assumptions
homo, normality w/ tx combo, independence, equal SS, sign interaction, tx’s are fixed or random
SPF assumptions
indenpendent, normal, homo of variance in tx a, sphercity holds, no interaction btwn tx groups & block (btwn factor)
ANCOVA assumptions
indepedence, Tx normal, Tx homogenous, x1 and y1 should be related, but not interacted, w/in group regress coefficients are equal, coviariance is w/o error, no interaction between tx & cov
are scores different across different subjucts?
CR-p
do scores of 1 subject differt between people?
RM
what are the effects of paper or electronic test scores when taking math, science or english?
CRF-pq
each person will take all tests in each format
RMF
Effects of paper or electronic testing while taking 3 different subjects?
SPF
