Sei sulla pagina 1di 4

FACTORIAL ANOVA

A factorial design is a way of thinking about research design as well as a way


of analyzing data.
Two-way ANOVA
A1
B

B1
B2

A
A2

A2B1
A1B1
A1B2 A2B2
MA1

A3
A3B
1
A3B
2
MA2

MB1
MB2
MA3

Factors - A & B; independent variables; experimental attribute.


Level (of factor) - A1 A2 A3 B1 B2; Values of independent variables.
Note: Table looks like X-tab but is not.
Questions: How is it different? How would you test whether the
independent variables (factors) were truly independent?
3 Questions of Interest (in a 2-way):
1. Effects of A?
1. and 2. are referred to as "Main Effects"
2. Effects of B?
3. Effects due to neither A or B alone, but attributable to the
combination of a particular
A group and B group. This is referred to as the "Interaction Effect"
(AxB).
Can be viewed as two separate experiments carried out on the same set of
subjects.
Can also test unique effects of combinations of treatments.
Completely crossed - each category or level of one factor occurs with each
level of every other factor.
Balanced - same number of cases in each cell.

Partition of the SS (two factor):


SStotal = SS factor A + SS factor B + SS AxB + SS within
Summary Table (2-way):
Source

SS

df

A (columns)

SSA

J-1

B (rows)

SSB

K-1

AxB (interaction)

SSAxB (J-1)(K-1)

Error (within cells)

SSerr

Total

SST

N-JK

MS
SSA/J-1
SSB/K-1

F
MSA/MSerr
MSB/MSerr

SSAxB
(J - 1)(K - 1) MS
AxB/MSerr
SSerr/(N-JK)

N-1

Compare with oneway ANOVA. We have now divided the SS into more
pieces.
Question: Where did the new pieces come from?
Answer: By controlling for the main effects of another variable and the
interaction effects, we have reduced error variance. Since error variance is
the "standard" against which we compare between group differences, we are
better able to discern differences we care about, i.e., we have increased
precision.
Recognizing an Interaction
F-test from ANOVA table
Profile plot from SPSS (lines not parallel)
Examine cell means (see Meyers, et al. handout)
Interpreting an Interaction
1.
2.
3.
4.

Select a variable to narrate


Narrate each level separately
Smoothly and grammatically integrate the sentences
Test for simple main effects (Multiple one-way ANOVAs simpler than
/lmatrix in Green and Salkind).

Effect Size for Factorial Anova:


From: http://jalt.org/test/bro_28.htm

Where:

Interpretation of these values is easiest if the decimal point is moved two


places to the right in each case, the result of which can be interpreted as
percentages of variance associated with each of the main effects, the
interaction, and error.

Where:

The interpretation of these partial eta2 values is similar to what we did above
for eta2 in that we need to move the decimal point two places to the right in
each case, and interpret the results as percentages of variance. However,
this time the results indicate the percentage of variance in each of the
effects (or interaction) and its associated error that is accounted for by that
effect (or interaction).
From Green and Salkind: Partial 2 ranges from 0 to 1. A partial 2 is
interpreted as the proportion of variance of the dependent variable that is
related to a particular main or interaction source, excluding the other main
and interaction sources. It is unclear what are small, medium and large
values for partial 2. What is a small versus a large 2 is dependent of the

area of investigation. In all likelihood, the conventional cutoffs of .01, .06,


and .14 for small, medium, and large 2 are too large for partial 2.
Specifying Fixed and Random Factors in Mixed Models (from Cornell
Statistical Consulting Unit)
Fixed factors can be thought of in terms of differences. The effect of a
categorical fixed factor is defined by differences from the overall mean and
the effect of a continuous factor is defined by its slope how the mean of the
dependent variable differs with alternate values of the factor. Conclusions
regarding fixed factors are particular to the values of these factors.
Random factors are defined by a distribution and not by differences. The
values of a random factor are assumed to be chosen from a population with
a normal distribution with a certain variance. The output for a random factor
is an estimate of this variance and not a set of differences from a mean.
Conclusions regarding random factors should be expressed in terms of
variance.
Situations that indicate fixed factors:
1. The factor is the primary treatment that the researcher wants to
compare.
2. The factor is a secondary covariate that might be confounded with the
treatment, and the researcher wants to control for differences in this
covariate. This would probably be better treated in SPSS by entering it
as a covariate rather than a factor.
3. The factor has only two values.
Situations that indicate random factors:
1. The researcher is interested in quantifying how much of the overall
variation to attribute to this factor.
2. The researcher is not interested in knowing which means differ, but
wants to account for the variation in this factor. Again, sounds like a
covariate.
3. The researcher would like to generalize the conclusions about this
factor to the whole population.
4.

Any interaction with a random factor is also random.

Potrebbero piacerti anche