Psych 7 Final Review

1
Psych 7 Final Review

• My name Aaron

2
Theory
Extraverts have low baseline levels of
physiological arousal
Extraverts will be high in sensation-seeking
Hypotheses
People who score highly on the NEO extraversion
scale will report stronger desire to try
sky-diving
Predictions
Data/Observations
Data from a specific sample do or do not support
the prediction
Theory is provisionally supported or
hypotheses/theory may require modification
Conclusions
3
Reliability

• Will a measure produce the same results
  consistently?
• Test-retest reliability (stability over time)
  correlation between first and second measurement
  of same individuals.
• Internal consistency reliability (how are
  different sub-components of the measure
  related?) correlation between scores on
  different scales of a measure.
• Split-half reliability correlation between
  scores on first half with those of second half.
• Cronbachs alpha average of inter-correlations
  across all items in a measure/scale.
• Inter-rater reliability (accuracy of an objective
  coding system) correlation between ratings made
  by different raters.

4
Construct Validity

• How well does a measure tap into the theoretical
  construct of interest?
• Face validity the measure satisfies intuition
  about the content of a construct.
• E.g., A measure of male genetic quality
• ?Fluctuating asymmetry mm deviations from
  perfect bilateral symmetry on traits like the
  ears, feet, and elbows.
• Not very face valid (i.e. not what you would
  picture when you think of genetic quality)
• But actually a theoretically compelling measure.

5
Construct Validity

• Predictive validity the extent to which scores
  on the measure predict (i.e. are correlated with)
  behaviors that should theoretically be related to
  the construct.
• E.g., Fluctuating asymmetry as a measure of
  genetic quality
• Does fluctuating asymmetry correlate negatively
  with the number of times men have been chosen as
  an extra-pair sexual partner?
• Yes. (Gangestad et al., 1997)

6
Construct Validity

• Concurrent validity (criterion groups validity)
  Do specific populations score on the measure as
  they would theoretically be predicted to?
• E.g., Fluctuating asymmetry
• Do male models have lower fluctuating asymmetry
  than a sample of men from a Star Trek convention?
• Note you are a weirdo if you actually collect
  this data.

7
Construct Validity

• Convergent validity Does a measure relate to
  other measures of a similar construct?
• E.g., does fluctuating asymmetry correlate
  negatively with male facial masculinity?
• Indeed, it does. (Gangestad Thornhill, 1998)
• Discriminant validity Is the measure
  (relatively) unrelated to measures of constructs
  you would theoretically predict it wouldnt be?
• E.g., does fluctuating asymmetry fail to
  correlate with female facial masculinity?
• Yes

8
Predicting Behavior behavior at the voting booth
Relationship (correlation) Between Variables
positive linear relationship
negative linear relationship
r -.90
r .90
curvilinear relationship
no relationship
r 0
9
Alternative vs. Clarifying Explanations
causes
causes
Ice cream
Cramps
Drowning
Temperature
Ice cream
Drowning
10
Basic Experimental Designs

• Posttest Only Design
• Pretest-Posttest Design
• Repeated Measures Design
• Matched Pairs Design

11
Dealing with Order Effects in Repeated Measures
Designs

• Complete Counterbalancing Present the different
  conditions in every possible order
• For 2 conditions (control/experimental) only 2
  possible orders
• For 4 conditions, 24 possible orders
• Randomized blocks Present conditions in random
  order
• Distracter Tasks/Time Lags Present one type of
  stimulus but then have rest period (avoid
  fatigue) or intervening task (e.g., count
  backwards) to avoid practice or contrast effects

12
The Latin Square

• In the Latin Square that appears above, four
  experimental conditions are represented by A, B,
  C, and D. The order in which subjects go through
  the conditions is represented by the sequence of
  letters in each row. Do the orders represented in
  this example validly implement the two rules for
  constructing a Latin Square?
• A) Yes each condition appears at each ordinal
  position and no rules are violated
• B) No, because the conditions are unequally
  represented in the third ordinal position
• C) No, because ABCD is a valid order but does not
  appear in the table
• D) b and c
• E) No, for a reason that is not listed in the
  above choices

13
Latin Square Designs

• Rule 1 Each condition appears at each ordinal
  position.
• Rule 2 Each condition precedes and follows each
  condition one (and only one!) time.

14
Factors vs Levels

• Factors variables which can contain numerous
  levels
• e.g., caffeine, sex
• Levels within each factor
• Caffeine level 1? 0 mg
• level 2? 50 mg
• level 3? 100 mg
• Sex level 1? male
• level 2? female

15
Main Effects

• Question How many main effects can there be for
  the factor caffeine (w/ 3 levels)?
• Caffeine level 1? 0 mg
• level 2? 50 mg
• level 3? 100 mg
• Answer 1

What kind of relationship could you test for
with this design that you couldnt with two
levels?
Answer A curvilinear/non-monotonic relationship
like an inverted U.
16
Alcohol Study in Factorial Notation
Expectation
Alcohol
No Alcohol
Alcohol
4.3
Actually Get
4.2
No Alcohol
5.05
3.45
17
Simple Main Effects in American Idol
Yeah!! Go Corey!
Type of Song
Overall Avg.
Love Song
Rock Song
150
Main Effect of Sex F M
Male
Sex of Singer
Female
250
DV of votes (in thousands)
18
Graphing Results of Factorial Designs
For Interaction Effects Plot all of the data as
line graphs. Nonparallel lines indicate an
interaction.
19
Comparing 2 Means

• Null hypothesis (H0) Population means are equal.
  Any differences between sample means are due to
  chance (random error remember, this is
  everything not in our manipulation).
• Research hypothesis (H1) Population means are
  not equal.
• T-test Test statistic associated with a
  probability of obtaining sample means that differ
  by observed amount if population means were equal
  (null hypothesis is true)

20
p-values

• Statistical significance assesses the
  probability that results could be due to chance
  rather than the hypothesized cause
• E.g., could difference between 2 means be as
  large as it is by chance?
• Where chance is everything not accounted for
  in your manipulation
• Another way to think of statistical significance
  is a measure of how likely it is that we have a
  true or real difference between groups

21
Phone numbers physics classes
Phone numbers no physics classes
t Difference between groups (means)
Normal variability within group(s)
22
Type I Type II Errors

• Type I error (a) incorrectly rejecting the null
  hypothesis when it is in fact correct (false
  positive)
• Men are designed to commit type I errors about
  female sexual desire.
• Male A Dude, did you see that? She totally
  wants me
• Male B Yeah. Sorry dude, but Im pretty sure I
  saw her clutching her can of mace
• Type II error (ß) incorrectly accepting the null
  hypothesis when it is in fact false (false
  negative)
• E.g., Lloyd and Harry at the end of Dumb
  Dumber
• Alpha is the p-value at which we decide to reject
  the null hypothesis
• As alpha gets larger, the probability of a type I
  error increases and the probability of a type II
  error decreases
• As N (sample size) increases, probability of a
  type II error decreases power (1- ß) increases

23

• Thats all, folks. Good Luck!
• Feel free to swing by Psych 2525 between 1 4
  today to ask additional questions.