You are on page 1of 93

Factorial Designs

Identifying and Interpreting Main Effects and Interactions

Factorial Designs

A simple experiment allows us to compare two conditions Reading times (ms) for a lexical decision task

Regularity Regular 500 Irregular 500

How would you compare the reading times for regular and irregular words?

Factorial Designs

But maybe the difference between regular and irregular words DEPENDS on whether the words are common (hi frequency) or uncommon (lo frequency) This is a factorial design and it allows us to ask more interesting questions

Frequency Hi

Regularity Regular Irregular 500 500

Lo

500

700

Factorial Designs

Factorial designs allow us to do two experiments at one time


One that compares regular to irregular words One that compares hi frequency to lo frequency words These are called main effects

Factorial Designs

Factorial designs also allow us to see if


The effect of frequency depends on whether the words are regular or irregular OR The effect of regularity depends on whether the words are hi or low frequency

These are called interactions

Factorial Designs

When looking at factorial designs, it helps to make a graph It is easier to see the effects if you use line graphs

Even if your really should be using bar graphs in the actual graphs that go in your paper

Factorial Designs

Graphing the Means

If there is an interaction between variables the lines are not parallel they have different slopes The DV always goes on the y-axis

One IV always goes on the x-axis


The other IV is plotted

Regularity Frequency Hi Lo Regular 500 500 Irregular 500 700

800 Irregular 700 Regular

600

500

400 Lo Hi

Factorial Designs

There is a main effect of frequency such that the responses were faster to hi frequency than lo frequency words

800 Irregular 700 Regular

600

500

400 Lo Hi

Factorial Designs

There is a main effect of regularity such that the responses were faster to regular words than irregular words

800 Irregular 700 Regular

600

500

400 Lo Hi

Factorial Designs

There is an interaction between frequency and regularity such that

for regular words there was no effect of frequency, however for irregular words responses were slower for lo frequency than hi frequency words for high frequency words there was no effect of regularity, however for low frequency words responses were faster for regular than irregular words

OR

Factorial Designs

There are always two ways to describe a two-way interaction Both are correct However, one often makes more sense than the other, or answers the research question better

Main Effects & Interactions

Two kinds of information can be gleaned from factorial designs Main Effects: An effect of a single IV

There is a main effect for each IV

Interactions: The effect of each IV across the levels of the other IV

The effect of one IV depends on the level of the other IV

Main Effects & Interactions

Main Effect

The main effect of each IV tells us about the relationship between that IV and the DV

Do different levels of an IV bring about different changes in the DV?

Need to look at row and column means

Main Effects & Interactions


Word Type Rehearsal Type Concrete Abstract

Rote
Imagery

5
10

5
5

Main Effects & Interactions


Word Type Rehearsal Type
Rote Imagery Column Means

Concrete
5 10

Abstract
5 5

Row Means

Main Effects & Interactions


Word Type Rehearsal Type
Rote Imagery Column Means

Concrete
5 10 7.5

Abstract
5 5 5

Row Means
5

7.5

Main Effects & Interactions


So what does this mean?

A main effect of Word Type tells us that more words are recalled when they are concrete A main effect of Rehearsal Type tells us that more words are recalled when imagery is used

For Example...
Regularity Frequency Hi Lo Mean

Regular 500 500

Irregular 500 700

Mean

Frequency

If you are talking about a main effect of frequency you are comparing hi frequency to lo frequency words PEROID. The word regularity should not appear in the sentence If you are talking about a main effect of regularity you are comparing regular to irregular words PEROID. The word frequency should not appear in the sentence

Regularity

Main Effects & Interactions

Is there an Interaction?

If so, then the main effects will have to be qualified, because an interaction indicates that the effect of one IV is different at different levels of the other IV

Main Effects & Interactions

Interactions in your everyday life

It depends indicates that what we do in one situation depends on some other variable For example: Whether or not you go to a party DEPENDS on whether you have to work and who is going to be there If you have to work you will not go If you do not have to work, you might go if a certain person is there

Main Effects & Interactions

To calculate interactions we are interested in differences If the differences are different then you have a two-way interaction
Regularity
Frequency Hi Lo Difference Regular 500 500 0 Irregular 500 700 200 Difference 0 200

Main Effects & Interactions


Word Type Rehearsal Type
Rote Imagery Difference

Concrete
5 10

Abstract
5 5

Difference

Main Effects & Interactions


Word Type Rehearsal Type
Rote Imagery Difference

Concrete
5 10 5

Abstract
5 5 0

Difference
0

Describing Main Effects and Interactions

In a 2x2 design there are THREE possible effects


A main effect of IV(A) A main effect of IV(B) A IV(A) x IV(B) interaction

You need to describe each in English

No Main Effect of Word Type No Main Effect of Rehearsal Type No Interaction


Rote 10 Imagery

# of Words Recalled

8 6 4 2 0 Abstract Concrete

Main Effect of Word Type (line is on a diagonal) No Main Effect of Rehearsal Type No Interaction
Rote 10 Imagery

# of Words Recalled

8 6 4 2 0 Abstract Concrete

No Main Effect of Word Type Main Effect of Rehearsal Type (space between lines) No Interaction
Rote 10 Imagery

# of Words Recalled

8 6 4 2 0 Abstract Concrete

Main Effect of Word Type Main Effect of Rehearsal Type No Interaction


Rote 10 Imagery

# of Words Recalled

8 6 4 2 0 Abstract Concrete

Main Effect of Word Type Main Effect of Rehearsal Type Interaction (lines are not parallel)
Rote 10 Imagery

# of Words Recalled

8 6 4 2 0 Abstract Concrete

Main Effect of Word Type No Main Effect of Rehearsal Type Interaction


Rote 10 Imagery

# of Words Recalled

8 6 4 2 0 Abstract Concrete

No Main Effect of Word Type Main Effect of Rehearsal Type Interaction


Rote 10 Imagery

# of Words Recalled

8 6 4 2 0 Abstract Concrete

No Main Effect of Word Type No Main Effect of Rehearsal Type Interaction


Rote 10 Imagery

# of Words Recalled

8 6 4 2 0 Abstract Concrete

Practice Makes Perfect!


For each of the following data sets:
1. 2. 3. 4.

Identify the IVs and their levels


Sketch a graph (by hand)

Calculate the main effects and interactions


Describe the main effects and interactions

Factorial Designs

Reaction Time (ms) to identify target Spatial Cue Gender Men Women Valid 500 500 Invalid 600 600

Men 700 650 600 550 500 450 400 Valid

Women

Invalid

Factorial Designs

Reading Times (ms) to identify target Luminance Frequency Lo Hi Lo 600 500 Hi 500 400

Lo Freq 700 650 600 550 500 450 400 350 Lo Lum

Hi Freq

Hi Lum

Factorial Designs

Reaction Time (ms) on Stroop Task Age Stimuli Consistent Inconsistent 6 yrs 1000 1100 18 yrs 600 900

Consistent 1550 1400 1250 1100 950 800 650 500 6yrs

Inconsistent

18 yrs

Factorial Designs

Recognition Accuracy (%) Visual Field Stimuli Words Faces Left 80 75 Right 90 65

Words 100

Faces

80

60

40 LVF RVF

See you Thursday!

Practice Makes Perfect!


For each of the following data sets:
1. 2. 3. 4.

Identify the IVs and their levels


Sketch a graph (by hand)

Calculate the main effects and interactions


Describe the main effects and interactions

Number of Words Recalled


Level of Processing
Sex Men Women Shallow 50 50 Deep 70 70

Number of Words Recalled


Level of Processing Sex Men Women Mean Diff Shallow 50 50 50 0 Deep 70 70 70 0 Mean 60 60 Diff 20 20

Number of Words Recalled

Men 100 Women

80

60

There is a main effect of levels of processing such that participants recalled more words with deep processing than with shallow processing

40 Shallow Deep

Reaction Time (ms)


Word Frequency Word Type Abstract
Concrete

Lo 800
900

Hi 400
500

Reaction Time (ms)


Word Frequency Word Type Abstract Concrete Mean Diff Lo 800 900 450 100 Hi 400 500 850 100 Mean 600 700 Diff 400 400

Reaction Time (ms)

Abstract 1000 800 600 400 200 0 Lo Hi Concrete

There is a main effect of word frequency such that participants were faster at processing hi frequency words than lo frequency words The is a main effect of word type such that participants were faster at processing abstract words than concrete words

Recognition Accuracy (%)


Visual Field
Stimuli Words Faces Left 40 60 Right 20 20

Recognition Accuracy (%)


Visual Field Stimuli Words Faces Left 40 60 Right 20 20 Mean 30 40 Diff 20 40

Mean Diff

50 20

20 0

Recognition Accuracy (%)

Words 100 80 60 40 20 0 LVF RVF

Faces

There is a main effect of stimulus type, such that participants were more accurate recognizing face stimuli than word stimuli The is a main effect of visual field, such that participants were more accurate recognizing stimuli in the LVF than in the RVF There is a significant interaction between stimulus type and visual field, such that in the LVF participants were more accurate recognizing face stimuli than visual stimuli, whereas in the RVF there was no difference in the recognition accuracy for face and word stimuli

Analyzing Factorial Designs


Getting to know and love SPSS

Analyzing Factorial Designs

In a 2x2 design, both factors can be


Both between-participants Both within-participants One between-participants and one withinparticipants

SPSS is different depending on the design

Analyzing Factorial Designs

We use SPSS to tell us if our main effects and interactions are significant SPSS is a good tool to support your analysis of what is going on in your data it should NOT drive your analysis

Analyzing Factorial Designs

Calculate means (the ANOVA will do this for you but just look at the means for now BY HAND

Make a 2x2 table of the means Calculate the main effects and interactions Draw graphs of the means

Describe the main effects and interaction in English Use SPSS ANOVA output to see if the main effects and interactions are significant

Analyzing Factorial Designs

A 2 x2 between-participants design

Randomized or Factorial

A 2 x2 between-participants and withinparticipants design

Mixed

A 2 x2 within-participants design

Repeated Measures

A 2 x2 between-participants design

LOP:

Deep Shallow

LOP
Stimulus Shallow Deep

Stimulus Type:

Visual Auditory

DV:

Auditory
Visual

Number of words recalled

Descriptive Stats

SPSS

A 2 x2 between-participants design
LOP
Auditory Visual

Stimuli Shallow

Deep

10 8 6

Auditory Visual

2.5 3.4

5.8 7.0

4 2 0 Shallow Deep

A 2 x2 between-participants design
LOP Stimuli Auditory Visual Mean Diff Shallow 2.5 3.4 Deep 5.8 7.0 Mean Diff

A 2 x2 between-participants design
LOP Stimuli Auditory Visual Mean Diff Shallow 2.5 3.4 2.95 .9 Deep 5.8 7.0 6.4 1.2

Mean
4.15 5.2

Diff
3.3 3.6

A 2 x2 between-participants design
Tests of Between-Subjects Effects Source ENCODE STIM Sum of Squares 119.025 11.025 df 1 1 Mean Square 119.025 11.025 F 231.616 21.454 Sig. .000 .000

ENC * STIM
Error

.225
18.500

1
36

.225
.514

.438

.512

Analyzing Factorial Designs

Two-way ANOVA

Indicates that there are two IVs Two main effects and one interaction

df main effects number of levels of the factor- 1 df interaction (A-1)(B-1) df error AB(n -1) F(1,28) = 13.95, p<.05

A 2 x2 between-participants design

The number of items remembered was analyzed in a 2 (encoding: shallow, deep) x 2 (stimulus: visual, auditory) factorial analysis of variance (ANOVA). There was a main effect of encoding, F (1,36) = 231.616, p < .001, such that recall was better with deep encoding (M = 6.400 , SD = .160) than with shallow encoding (M = 2.590, SD = .160). There was a main effect of stimulus, F (1, 36) = 21.454, p < .001, such that recall was better for visual (M = 5.200 , SD = .160) than for auditory stimuli (M = 4.150 , SD = .0160). There was no significant interaction between encoding and stimulus, F (1, 36) = .436, p > .05.

A 2 x2 mixed design

Sex:

Men Women

Sex
Attention Men Women

Attention:

Focused Divided

DV:

Focused
Divided

Number of words recalled

Descriptive Stats

SPSS

A 2 x2 mixed design

Sex
Men Women

Attention

Men

Women

10 8 6 4

Focused Divided

7.06 4.00

7.44
2

3.56

0 Focused Divided

A 2 x2 mixed design
Sex Attention Focused Divided Mean Diff Men 7.06 4.00 Women 7.44 3.56

Mean

Diff

A 2 x2 mixed design
Sex Attention Focused Divided Mean Diff Men 7.06 4.00 5.53 3.06 Women 7.44 3.56 5.50 3.88

Mean
7.25 3.78

Diff .38 .44

A 2 x2 mixed design
Tests of Within-Subjects Effects Source Sum of Squares ATTEN Sphericity Assumed 192.516 Greenhouse-Geisser 192.516 Huynh-Feldt 192.516 Lower-bound 192.516 ATTEN * SEX Sphericity Assumed 2.641 Greenhouse-Geisser 2.641 Huynh-Feldt 2.641 Lower-bound 2.641 Error(ATTEN) Sphericity Assumed 46.344 Greenhouse-Geisser 46.344 Huynh-Feldt 46.344 Lower-bound 46.344 df Mean Square F Sig.

1 1.000 1.000 1.000


1 1.000 1.000 1.000 30 30.000 30.000 30.000

192.516 192.516 192.516 192.516


2.641 2.641 2.641 2.641 1.545 1.545 1.545 1.545

124.622 124.622 124.622 124.622


1.709 1.709 1.709 1.709

.000 .000 .000 .000


.201 .201 .201 .201

A 2 x2 mixed design
Tests of Between-Subj ects Effects Measure: MEASURE_1 Transformed Variable: Average Source Intercept SEX Error Type III Sum of Squares 1947.016 .016 40.469 df 1 1 30 Mean Square 1947.016 .016 1.349 F 1443.347 .012 Sig. .000 .915

A 2 x2 mixed design

Accuracies were analyzed in a 2 (sex: men, women) x 2 (attention: focused, divided) mixed analysis of variance (ANOVA). There was a main effect of attention, F (1, 30) = 124.622, p < .001, such that accuracy was better under focused attention conditions (M = 7.250 , SD = .211) than under divided attention conditions (M = 3.781, SD = .214). There was no main effect of sex, F (1, 30) = .012, p > .05, nor was there a sex by attention interaction, F (1, 30) = 1.709, p > .05.

A 2 x2 mixed design

Frequency:

Lo Hi

Frequency
Regularity Lo Hi

Regularity:

Regular Irregular

DV:

Regular
Irregular

Reaction Time (ms)

Descriptive Stats

SPSS

A 2 x2 mixed design

Frequency
Regular Irregular

Regularity

Lo

Hi

700

500

Regular Irregular

572 699

540 544
300 Lo Hi

A 2 x2 mixed design
Frequency Regularity Lo Hi Mean Diff

Regular Irregular

572 699

540 544

Mean Diff

A 2 x2 mixed design
Frequency

Regularity

Lo

Hi

Mean

Diff

Regular Irregular

572 699

540 544

556
622

32
155

Mean Diff

636 127

542 4

Tests of Within-Subj ects Effects Measure: MEASURE_1 Source FREQ Type III Sum of Squares 68578.516 68578.516 68578.516 68578.516 19117.234 19117.234 19117.234 19117.234 139782.516 139782.516 139782.516 139782.516 11000.234 11000.234 11000.234 11000.234 60823.891 60823.891 60823.891 60823.891 14328.859 14328.859 14328.859 14328.859 df 1 1.000 1.000 1.000 15 15.000 15.000 15.000 1 1.000 1.000 1.000 15 15.000 15.000 15.000 1 1.000 1.000 1.000 15 15.000 15.000 15.000 Mean Square 68578.516 68578.516 68578.516 68578.516 1274.482 1274.482 1274.482 1274.482 139782.516 139782.516 139782.516 139782.516 733.349 733.349 733.349 733.349 60823.891 60823.891 60823.891 60823.891 955.257 955.257 955.257 955.257 F 53.809 53.809 53.809 53.809 Sig. .000 .000 .000 .000

Error(FREQ)

REG

Error(REG)

FREQ * REG

Error(FREQ*REG)

Sphericity Assumed Greenhouse-Geisser Huynh-Feldt Lower-bound Sphericity Assumed Greenhouse-Geisser Huynh-Feldt Lower-bound Sphericity Assumed Greenhouse-Geisser Huynh-Feldt Lower-bound Sphericity Assumed Greenhouse-Geisser Huynh-Feldt Lower-bound Sphericity Assumed Greenhouse-Geisser Huynh-Feldt Lower-bound Sphericity Assumed Greenhouse-Geisser Huynh-Feldt Lower-bound

190.608 190.608 190.608 190.608

.000 .000 .000 .000

63.673 63.673 63.673 63.673

.000 .000 .000 .000

A 2 x2 within-participants design

Response times were analyzed in a 2 (frequency: hi, lo) x 2 (regularity: regular, irregular) repeated-measure analysis of variance (ANOVA). There was a main effect of regularity, F (1, 15) = 190.608, p<.001, such that responses were faster to regular words (M = 542.594 , SD = 24.610) than irregular words (M = 636.063 , SD = 23.638). There was a main effect of frequency, F (1, 15) = 53.809, p<.001, such that responses were faster to hi frequency words (M = 556.594 , SD = 23.787) than to low frequency words (M = 622.063 , SD = 24.810). There was a significant frequency by regularity interaction, F (1, 15) = 63.673, p<.001, such that there was an effect of frequency for irregular words, but not regular words.

Writing the Results

Writing the Results

In the results section you are presenting the findings of your experiment A good pattern is to report the results in statistical language, followed by a statement in English about what that means

Some experiments require you to do more than one set of analyses put each set in a separate paragraph

Writing the Results

All results sections should begin with a statement about how you reduced the data, and then refer to a table or figure where you present the data itself For example, in a typical RT experiment there are many trials, but those are reduced to the means for each condition for each subject

Did you eliminate any subjects at this stage for having error rates that were too high or other reasons that make their data suspicious? Report them here. Present the actual data in a table OR figure

Results: Tables OR Figures


Tables OR figures help clarify the results Generally, tables are used to present large arrays of data (15+ means) In the text, refer to a table or figure by # and describe As shown in Figure 2, the aerobics group

Tables or figures supplement the text, they do NOT replace it

Writing the Results

There are no standards on the reporting of statistics


Is there a difference Stats to back up difference How were they different

I would like you to report exact p values, to 3 decimal places If SPSS tells you that p = .000, report that p< .001

t-Test (independent or paired)


Start with a description of your data Report the results of the t-test, followed by an English statement of which mean was the higher

A significant t-test tells you that two means are different, it doesnt tell you which one was higher
E.g., Number of items recalled in each encoding condition was compared with an independent t-test. There was a significant difference between conditions, t(32) = 2.95, p = .03. More words were recalled in the semantic encoding condition (M = 16, SD = 1.4) than in the phonological encoding condition (M = 12, SD = 1.2).

ANOVA

There are many types of ANOVAs, but they all have the same basic format:

There are 2 or more factors (independent variables), each of which has 2 or more level The factors can be either within-subjects or betweensubjects

Start with a statement about how you prepared the data for analysis
Present the data, either in a table OR a figure

E.g., Mean response times were calculated for each condition, and are presented in Table 1.

ANOVA

Introduce your ANOVA and present your design


Mention each factor, and the levels of each factor If all of your factors are b/t, you can call it a factorial ANOVA If all of your factors are w/in, you can call it a repeated measures ANOVA If you have some of each, you call it a mixed ANOVA, and then specify which factors are w/in and which are b/t E.g., Response times were analyzed in a 2 (encoding: shallow, deep) x 2 (modality: auditory, visual) mixed Analysis of Variance (ANOVA), with encoding as a withinsubjects factor and modality as a between-subjects factor.

ANOVA

Report the main effects, one at a time In a very complex design (e.g., in a 2 x 2 x 2 x 3 design there are 4 main effects, 6 2-way interactions, 4 3-way interactions, and a 4-way interaction) you might report only the significant main effects, and the theoreticallyinteresting non-significant effects

However, since you will be doing only very simple designs, report ALL of your main effects and all of your interactions, significant or not

ANOVA

When you report the effect, first describe the effect in statistics, then in English (or, if you can combine them)

E.g., There was a main effect of gender, F(1, 23) = 3.16, p = .022, such that women were funnier than men.

OR

E.g., Women were funnier than men, F(1, 23) = 3.16, p = .022

ANOVA

If the interaction is significant, you need to check to see if the main effect is still valid (sometimes it is, but sometimes it isnt) If the main effect is misleading (i.e., the effect holds for one level but not the other), you need to qualify it, so that your reader knows not to be fooled

E.g., There was a main effect of regularity, F(1, 31) = 5.67, p = .01, that was qualified by the frequency x regularity interaction (then you would go on to describe the interaction)

ANOVA

Describe the interaction If it is NOT significant just say that its not significant, and report the F (you can report the exact p, or you can report ns, which stands for not significant) If it IS significant, report it, and then describe it in English

E.g., There was an interaction between word frequency and regularity, F(1, 31) = 5.67, p = .008. For high frequency words, response times were the same for regular and irregular words however, for low frequency words, response times were greater for irregular than for regular words.

ANOVA

Sometimes an interaction occurs when both levels show the same pattern of results, but the effect is greater for one than the other

E.g., There was an interaction between word frequency and regularity, F(1, 31) = 5.67, p = .008, such that irregular words produced greater slowing for low frequency words, than for high frequency words.

Assignment 3

Assignment #3

You will be given the data files for four questions For each question, read the experiment description that I provide, analyze the data in SPSS, and write a one-paragraph results section Your submitted assignment should consist of:

Title Page Results sections, each on a separate page Tables (you need to put the data in a Table if you arent going to put the means in the text) Figure captions Figures, each on a separate page

NOTE: You dont need Tables or Figures for experiments you can analyze with a t-test, but you will need it for other experiments.

You might also like