Bi-Modal Scales--Middling Liability What is your view on gun control? Totally against -3 Mainly...
-
Upload
emerson-bickford -
Category
Documents
-
view
216 -
download
1
Transcript of Bi-Modal Scales--Middling Liability What is your view on gun control? Totally against -3 Mainly...
Bi-Modal Scales--Middling Liability
What is your view on gun control?
Totally against
-3
Mainly against
-2
Some-what
against
-1
Neither for nor against
0
Some-what favor
1
Mainly favor
2
Totally favor
3
1. Equal numeric values ≠ equal psychological values
2. “0” = undecided? Ambivalent? Unfamiliar? Don’t Care?
Measure Attitude Behavior
Support for Public Schools
Survey
Degree values public
schools
Votes to increase local taxes for
schools.
IMPLICIT MODEL OF ATTITUDE MEASURES
Problems with Verbal Measures Response Biases:
1. Social desirability2. Sabotaging
Affected by situations and contexts Salience problem: Ss know that they are being measured.
a. High salience attn, but biasb. Low salience attn but accuracy
Reactivity problem: IV = (IV + Measure) May require a-typical depth of processing/introspection Can assume people know own inner states/inner processes more than they actually do.
Information Processing Model of Survey Response Strack & Martin, 1987
Moral Values InventoryRettig & Pasamanick, 1959
Sample items to display inter-generational limitations of empirical realization
TO WHAT DEGREE WOULD YOU CONDONE:
Item 6: Girls smoking cigarettes
Item 31: Buying bootleg liquor under prohibition law
Item 39: Seeking amusement on Sunday instead of going to church.
Bi-Modal Scales--Middling Liability
What is your view on gun control?
Totally against
-3
Mainly against
-2
Some-what
against
-1
Neither for nor against
0
Some-what favor
1
Mainly favor
2
Totally favor
3
1. Equal numeric values ≠ equal psychological values
2. “0” = Undecided? Ambivalent? Unfamiliar? Don’t Care?
Behavioral Measures
1. Overt behavior 2. Behavioroid
3. Physiological
Advantages of Behavioral Measures
1. More absorbing
2. Require less inference of rel. btwn IV and behavior, b/c measure IS behavior.
3. Tells a better story
Types of Behavioral Measures
Frequency Extent/Amount Speed Intensity Duration Preference Latency Social/Physical Distance Non-verbal Cues and
Expressive Behaviors
Unobtrusive Measures
Non-Verbal Behavior as DV
Non-Verbal Behavior as DV
Behavioroid Measures
Defined: Measure INTENT to commit the behavior, w/o actually measuring or inducing behavior.
Used when actual behavior is too impractical, unethical, or otherwise inappropriate.
Example of Behavioroid Measure
Freedman and Fraser "Foot in the Door" StudyJPSP, 1966
Behavioroid Measure:
Willingness to have 2.5 hour intrusive survey of house conducted by 5 strangers.
a. Not previously contacted: 22.2%b. Familiarized with survey questions: 27.8%c. “Complete short survey?”, not administered 33.3%d. “Complete short survey?”, administered 52.8%
Physiological Measures
Defined: Bodily states that reflect psychological states
Examples: Blood pressure, heart rate, skin conductance
Advantages:
Problems:
Not under conscious control
Display mediation
Costly; Intimidating to subjects;
Typically gross, rather than subtle;
Require inference to conceptual DV
Indirect Measures
Measures that imply DV, without directly testing it.
Observable Behavior Implied State
Physical Distance from a minority person
Reselling price for chosen vs. given item.
Eye contact during "get acquainted" meeting.
Hostility
Valuation due to perceived control
Liking, attraction
Concluding Points Re. DVs1. Which is the better feedback bias measure?
a. Feedback bias = overall rating of essay
b. Feedback bias = (# pos. comments − # neg. comments)
2. Which is better measure of hostility to out-group?
a. Amount of shock delivered during “learning” task
b. Physical distance during interview
Aim for actual and behavioral, rather than general attitude
Expt. DV should be close to conceptual DV.
3. Which is better measure of health after disclosure?
a. Visit vs. Did Not Visit MD
b. Number of MD visitsDV should be as precise and sensitive as possible
Experimental Designs
Class 10
C = Control Cond
B = Black Writer Cond
W = White Writer Cond
IV Induces, DV Confirms
Reliability and Validity of Dependent Variables
Reliability: DV provides consistent measurement
Temporal: Test/Re-test
Inter-observer or Inter-rater
Inter-item reliability (for scale development)
SPSS Reliability Output
Inter-Item Reliability for Optimism Measure
AKA “Cronbach’s Alpha”
LOT = Life Orientation Test
Scheier, Carver, & Bridges, 1994
Measures of Reliability
“Mood” = sad + angry + (not) happy + afraid
Reliability Type
Internal consistency (Cronbach’s Alpha)
Rater A and Rater B agree 92% Inter-rater
Ethics measure taken at start of term agree with same measure taken end of term
Test-retest
Method
ValidityValidity: DV measures what it is supposed to measure
Face ValidityThe measure has a “common sense” resemblance to the
construct
Criterion-Related Validity
The measure is confirmed by a more rigorous standardConcordant validityPredictive validity
Construct Validity
Proves merit of underlying constructBest obtained through multiple measures
Face Validity
Construct
Hostility
Prejudice
Need for affiliation
Attraction
Empirical Realization
No. of shocks to harasser
RT: see Black or White target, ID pos/neg words
Choosing to be alone or with another person
Pupil dilation
Face Valid?
(High, Med., Low)
High
Med/low
High
Low
Criterion Validity
Match or fit between specific, empirical DV (i.e., the one used in study) and an independent (and presumably stable, encompassing) measure of conceptual DV (i.e., the criterion).
Concordant validity: Criterion exists in the present.
Predictive validity: Criterion represented by future behavior.
Expt’l DV Criterion
Gym visits Treadmill Endurance
Conscientiousness No. of missed classes Survey (week 1) (end of semester)
Validity Type
Concordant
Predictive
Conceptual DV
Fitness
Reliability
Construct Validity
Face validity and criterion validity refer largely to the validity of the measure.
Construct validity refers to the validity of the underlying conceptual DV.
Typically requires multiple measures
Convergent Validity
Different measures that have only the underlying construct in common.
Neuroticism: moderately related to stress, negative affect, self-preoccupation, fear of judgment
Divergent Validity
Measure is not tightly related to similar constructs
Neuroticism introversion, conscientiousness
Neuroticism
Validity
Sexism scale items include:Women demand too many rights Wives should vote as their husbands do
Validity Type
Face Validity
“Acrophobia Survey” verified with heart rate, sweating, hyperventilation
Criterion
Aggressiveness Measure taken at prescreening predicts shocks delivered in experiment, 3 weeks later
Predictive
Method
Validity (continued)
“Nurturance” = Attn. to other’s emotions Listening to problems Willing to help
Validity Type
Construct Validity
Self Esteem moderately related to:
Self-confidence; Self-Clarity; Self- acceptance
Convergent Validity
Method
Self Esteem is not highly correlated with self-efficacy
Divergent Validity
Boosting Validity
Avoid “response set”: Alternate (or mix) the positive and negative valence of questions, in survey DV.
Systematic Replication: Several experiments, each one accounts for alternative explanation
Disguise measure: “Chicken game” in Culture of Honorstudies—non obvious measure of aggression.
DV outside of conscious control, e.g., physio reactions
Weighing the Alternatives
I had to shed 20 pounds or else I’d lose my job and my wife would leave me and I’d die an early death. I was desperate for a solution. Then I found Chubby Checkers ®. Within 2 months I lost 15
pounds! You can, too!
Implied causal story?
Alternative explanations?
Saturday Academy Research Design
Pre-test SAT= 940
Class sessions
Post-testSAT = 991
SAT gain pre to post test = 51 points
p < .01
Implied causal story?
Alternative explanations?
What does design need to address alternative explanations?
Control Groups
Purpose: To establish causality; that it is IV, and only IV, that accounts for DV.
Attributes of Control Group:
1. Random selection: * Each participant is equally likely to be
assigned to expt'l or control condition. * Provides a check on systematic error
But, does not control for random error
2. Control condition should mimic experimental condition in all respects other than the IV.
3. Assign Ss to control or experimental conds. just before
introducing IV
Counterbalancing
Sub. No.
1
2
3
4
5
6
7
8
Confederate Race
Black
White
Black
White
Black
White
Black
White
Confederate Temperament
Friendly
Friendly
Unfriendly
Unfriendly
Friendly
Friendly
Unfriendly
Unfriendly
Essay Topic
TV
TV
TV
TV
Environ
Environ
Environ
Environ
Single Factor Design
(1 X 2) (From Interracial Feedback Research)
Black White
Writer Race I II
1 Factor: Race of Writer
2 Levels: Black writer or White writer
Shows main effect only (whether phenomenon exists).
Does not show interaction.
Components of Factorial Design
Factors: The independent variables
Factor 1: Writer raceFactor 2: Writer friendliness
‘
Levels: The dimensions within factors
Level 1, Friendliness: FriendlyLevel 2, Friendliness: Unfriendly
Conditions: The intersection of factors and levels
Condition I: Friendly, Black writerCondition IV: Unfriendly, White writer
Black White
Friendly I II
Unfriendly III IV
2 X 2 Friendly Unfrnd
Black Ia 1b
White 2a 2b
2 X 3 Friendly Unfrnd Neutral
Black 1a 1b 1c
White 2a 2b 2c
3 X 3 Friendly Unfrnd Neutral
Black 1a 1b 1c
White 2a 2b 2c
Asian 3a 3b 3c
Friendly Unfrnd
Black Ia 1b
White 2a 2b
Friendly Unfrnd
Black Ia 1b
White 2a 2b
2 X 2 X 2Race-relevant Essay Race-Irrelevant Essay
Factorial Designs as Coherent Sentences
Number of Factors
Factorial “Sentence”
1 Ethical decisions (blind/don/t blind) are affected by discussion opportunity (discuss vs. don’t discuss).
2 Ethical decisions are affected by group discussion and social contexts (Ghakistan vs. NY)
3 Ethical decisions are affected by group discussion and by social context as a function of gender.
4 Ethical decisions are affected by group discussion and by social context, as a function of gender—but only among college educated.
5 Ethical decisions are affected by group discussion and by social context, as a function of gender—but only among the college educated, who specialized in humanities rather than engineering.
Determining Number of Levels w/n Factors
How does arousal affect test performance?
0
2
4
6
8
10
Low Arousal High Arousal
02468
10
LowArousal
Mod.Arousal
HighArousal
Yerkes-Dotson Law
Perfo
rman
ce L
evel
Low Moderate High
Arousal Level