JEFF'S PSYCH 210 EXAM 4 JMU STUDY GUIDE
Analysis of Variance (ANOVA) - Answer -tests whether 3 or more groups' means are
different (is there a significant difference among the groups)
one-way ANOVA - Answer -1 IV with three or more levels (multi-group). between or
within subjects design
two-way ANOVA - Answer -two IVs with two or more levels (factorial). between, within,
or mixed subjects design
Why not use multiple t-tests? - Answer -1) could take a while
2) the alphas are cumulative across the experiment, so the more tests you do, the more
error there is
within-group variation (error) - Answer -if the null is true, any numerical differences will
be "within" each group since the samples came from the same population. therefore,
any numerical differences are error
between-group variation (sample means) - Answer -if the null is NOT true, then
between-group variation is due to change in the IV
F - Answer -F = between-group variance divided by within-group variance OR
MSbetween divided by MSwithin. you want F to be big. if null is true, F should be 1
sources of variance - Answer -1) measurement error (DV) [error]
2) uncontrolled variables [error]
3) participants [error in between-subject, not in within]
4) the IV
total variance for one-way between-subjects ANOVA - Answer -made up of error and
IV, want IV to make up biggest part
total = between + within
ANOVA vs post hoc - Answer -ANOVA: is there a difference somewhere?
Post hoc: where is the specific difference?
design vs statistics - Answer -design: between/within/mixed-subjects multi-
group/factorial
statistics: one/two-way between/within/mixed-subjects ANOVA
F-statistic hypothesis testing - Answer -1) define null (no difference in means of groups)
and alternative hypothesis, choose alpha
2) determine MSbetween and MSwithin and df
3) calculate F statistic
, 4) compare Fcalc with Fcrit (based on alpha, dfbetween, and dfwithin). if Fcalc≥Fcrit,
reject null
5) interpret your conclusions (there is/isn't a difference between groups)
6) conduct post-hoc test to determine which groups differ significantly
ANOVA table - Answer -how the data is presented
identifying ANOVA tables - Answer -based on the number of parts that variance is
broken up into (one-way between has variation in two parts, one-way within has
variation in three parts, two-way between in four parts)
determining significance in ANOVA - Answer -compare Fcalc to Fcrit
Fcrit found in F table based on the between and within degrees of freedom
Fcalc ≥ Fcrit, reject null, p<.05
conclude there is a significant difference somewhere among the groups
Effect Size for one-way between - Answer -the percent of variance explained.
SSbetween is what we can explain because it came from our manipulation
Post Hoc Testing - Answer -tells exactly where the differences between groups exist
Issues with Post Hoc Testing - Answer -Tests vary from conservative (needs a big
difference to be conservative) to liberal (small differences can be significant). Also need
to control for experiment-wise error rate
common report of ANOVA findings - Answer -F(between df, within df, MSE) = Fcalc ,
p</> .05
Tukey's Post Hoc - Answer -about 65% liberal and controls for error
Mean Square Error (MSE) - Answer -mean square of the within variation
MS equation for everything - Answer -SSwhatever divided by dfwhatever (whatever
you're trying to find, between, within, etc)
SStotal equation - Answer -Σ(X-GM)^2
SSbetween equation - Answer -Σ(M-GM)^2
SSwithin equation - Answer -Σ(X-M)^2 (M is of cells in factorial) (THIS IS ERROR)
SStotal full equation - Answer -Σ(X-GM)^2 = Σ(M-GM)^2 + Σ(X-M)^2
SSrows equation - Answer -Σ(Mrows-GM)^2
SScolumns equation - Answer -Σ(Mcolumns-GM)^2
Analysis of Variance (ANOVA) - Answer -tests whether 3 or more groups' means are
different (is there a significant difference among the groups)
one-way ANOVA - Answer -1 IV with three or more levels (multi-group). between or
within subjects design
two-way ANOVA - Answer -two IVs with two or more levels (factorial). between, within,
or mixed subjects design
Why not use multiple t-tests? - Answer -1) could take a while
2) the alphas are cumulative across the experiment, so the more tests you do, the more
error there is
within-group variation (error) - Answer -if the null is true, any numerical differences will
be "within" each group since the samples came from the same population. therefore,
any numerical differences are error
between-group variation (sample means) - Answer -if the null is NOT true, then
between-group variation is due to change in the IV
F - Answer -F = between-group variance divided by within-group variance OR
MSbetween divided by MSwithin. you want F to be big. if null is true, F should be 1
sources of variance - Answer -1) measurement error (DV) [error]
2) uncontrolled variables [error]
3) participants [error in between-subject, not in within]
4) the IV
total variance for one-way between-subjects ANOVA - Answer -made up of error and
IV, want IV to make up biggest part
total = between + within
ANOVA vs post hoc - Answer -ANOVA: is there a difference somewhere?
Post hoc: where is the specific difference?
design vs statistics - Answer -design: between/within/mixed-subjects multi-
group/factorial
statistics: one/two-way between/within/mixed-subjects ANOVA
F-statistic hypothesis testing - Answer -1) define null (no difference in means of groups)
and alternative hypothesis, choose alpha
2) determine MSbetween and MSwithin and df
3) calculate F statistic
, 4) compare Fcalc with Fcrit (based on alpha, dfbetween, and dfwithin). if Fcalc≥Fcrit,
reject null
5) interpret your conclusions (there is/isn't a difference between groups)
6) conduct post-hoc test to determine which groups differ significantly
ANOVA table - Answer -how the data is presented
identifying ANOVA tables - Answer -based on the number of parts that variance is
broken up into (one-way between has variation in two parts, one-way within has
variation in three parts, two-way between in four parts)
determining significance in ANOVA - Answer -compare Fcalc to Fcrit
Fcrit found in F table based on the between and within degrees of freedom
Fcalc ≥ Fcrit, reject null, p<.05
conclude there is a significant difference somewhere among the groups
Effect Size for one-way between - Answer -the percent of variance explained.
SSbetween is what we can explain because it came from our manipulation
Post Hoc Testing - Answer -tells exactly where the differences between groups exist
Issues with Post Hoc Testing - Answer -Tests vary from conservative (needs a big
difference to be conservative) to liberal (small differences can be significant). Also need
to control for experiment-wise error rate
common report of ANOVA findings - Answer -F(between df, within df, MSE) = Fcalc ,
p</> .05
Tukey's Post Hoc - Answer -about 65% liberal and controls for error
Mean Square Error (MSE) - Answer -mean square of the within variation
MS equation for everything - Answer -SSwhatever divided by dfwhatever (whatever
you're trying to find, between, within, etc)
SStotal equation - Answer -Σ(X-GM)^2
SSbetween equation - Answer -Σ(M-GM)^2
SSwithin equation - Answer -Σ(X-M)^2 (M is of cells in factorial) (THIS IS ERROR)
SStotal full equation - Answer -Σ(X-GM)^2 = Σ(M-GM)^2 + Σ(X-M)^2
SSrows equation - Answer -Σ(Mrows-GM)^2
SScolumns equation - Answer -Σ(Mcolumns-GM)^2
