
Correlational Research (4)
 Examining relationships between and among variables
 Seeing if two variables are related to each other
 Useful for
  description
  manipulation
 Extent to which 2 measured variables are sytematically related

Why can't correlations establish causality? (2)
 Dirction of casuality
 X could cause Y
  or Y could cause X
 3^{rd }Variable Problem
  something else could cause both X and Y

Correlational Coefficient (r) (3)
 statistic that shows the direction of a relationship between 2 variables
 show the direction of a relationship: postivite/negative
 shows the magnitude (strength) of a relationship: relatively weak/relatively strong

Positive Correlations
 As one goes up, the other goes down
 go in the same direction
 direct relationship

Negative Correlations
 Indicated by absolute value:
 closer to 1 (1) = stronger
 closer to 0 = weaker
 strong relationship: variables are more related
 weak relationship: variables are less related

The statistical signficance of r is affected by what 3 factors?
 sample size
 # of participants min. r for sig.
 >>10_________.52
 >>100________.16
 >>1000_______.05
 magnitude of correlation: larger r is more likely to be significant
 conventional rule
 >> .10 = weak correlation
 >> .30 = moderate correlation
 >> .50 = stronge correlation
alpha level: how sure we want to be

2 Aspects of Experimental Power
 manipulation: making something happen
 control: keeping everything the same except the manipulated variable

5 Aspects of True Experiments
 IV is manipulated
 quantitatively (including high, low & no)
 qualitatively (e.g., rehearsal vs imagery)
random assignment
extraneous factors are controlled
DV is measured
Basic logic: if DV varies as a result of manipulation of the the IV, than the IV caused the DV to change.

Internal Validity (3)
the basic minimum without which any experiment is uniterpretable
degree to which you can confidently draw a casual inference from the data (i.e., the IV caused the DV to vary)
 degree to which your study is free from confounds

Mills' 3 Requirements for Establishing Casuuality
 Covariation: 2 things go together (change at the same time)
 Temportal Sequence: 1 thing precedes another
 Eliminating Confounds: other things that didn't cause it to happen

Confounding Variables (confounds, nuisance variables) (2)
any uncontrolled extraneous variable that "covaries" with the independent variable and could provide an alternative explanation of the results
literally means mistaking one thing for another

3 Causes of Confounds
Participant/subject variables: something about the subjects in the study (e.g., partcipants are selected base on the same trait).
Environmental variables: something about the setting the subjects are in.
Procedural variables: something about the procedure  the way subjects are treated.

3 Ways to Deal with Confounds
 Be careful
 read as much as possible about your topic to reduce the chance of making mistakes
 Replicate Your Findings
 use converging operations to study the same topic
 Measure and Control
 if you can, control it from the start
 if you can't, measure it and control it statistically

3 Categories of Participant Confounds
 "People are different"
 individual differences can undermine careful, systematic scientific observation (e.g., selection bias)
 "People change"
 differences within people across time and situations can masquerade as meaningful research findings (e.g., history, maturation and regression toward the mean)
 "The process of studying people changes people"
 many studies have the uninteneded consequence of changing people's behavior in some small way (e.g., tesing effects, mortality, reaction biases and instrumentation)

Selection Bias (2)
sampling people from an unrepresentative sample
 2 kinds
 selfselection
 selection of samples that may differ to begin with

History
changes that occur in a very larege group of people such as a nation of culture

Maturation
specific developmental or experimental changes that occur in a particular person, or a particular cohort overtime

Regression Toward the Mean
 extreme high/low scores tend to move closer to the mean
 may occur when participants have been selected from a study based on extreme scores

Testing Effects
changes in test scores are due to repeated testing rather than manpulation of the IV

Mortality (Attrition)
 participants are lost from the study
  group differences may be due to differences in those remaining and not to the manipulation of the IV

Participant Reaction Bias
 people behave differently when they know they are being studied

3 participant expectancies
 demand charactereistics: give hints as to what the experimenter wants
 participant reactence: participants rebel against an experimental manipulation
 evaluation apprehension: people are concered about being judged

Experimenter Bias (aka experimenter expectancy) (2)
errors in a study due to the predisposed notions or beliefs of the experimenter
unintentional behaviors on the part of the experimenter

Instrumentation
 changes in the DV are produced by changes in the measurement instrument itself
  a bad scale (unreliable)
  a bad measure (rater)
  time of measurement

4 Ways to Minimize Confounding Variables
 Use random assignment
 law of large numbers
 most things are normally distributed
Pretesting: can check baseline differences and control is necessary
Matching: ensures confounding variables are evenly distributed (e.g., gender, IQ)
Experimental Control: keep/hold things constant

Testing vs Instrumentation
both deal with data collection
testing sensitizes participants to data collection process
instrumentation changes the data collection process in some way

Solving Thereats to Internal Validity (3)
 comparison/control group
 random assignment
 experimental control
*need to think about threats while in the design phase of a study

Manipulating the IV (3)
operationally define the construct (concept)
provide informed consent
explain without revealing the hypothesis

2 Types of Manipulation
 straightforward: no deception involved (only type used in psy 2200)
 staged: set up "real situational/psychological state (often uses deception/confederates)

Strength of Manipulation
 Weak: run risk of not detecting small, but meaningful, differences between groups
 Strong: maximizes differences between groups and may not reflect reality

3 Types of IVs
 Situational: features in the environment
  # of people present
  level of lighting, volume of music
 Task Variables: different kinds of problems to solve
  abstract v concrete
  easy v difficult
 Instructional Variables: different ways to perform a task
  imagery v rehearsal
  argue for/against your own position

Central Limit Theorem
 the distribution of means will increasingly approximate a normal distribution as the size N of samples increases
 i.e., the larger the sample size the more normal the distribution will be

Random Sampling/Selection (2)
the method of selecting a sample from a population
every memeber of the population has an equal chance of being selected for the study

Random Assignment (2)
the method of allocating participants to groups
each participant is as likely to be assigned to one group as to another

Extraneous Variables
any thing other than the manipulated variable that may affect the outcome (aka control variables/nuisance variables)

Measures are based on (3)
 theory
 past research
 intuition

3 Types of Measures
 selfreport
 behavioral (covert, overt)
 physiological responses

2 Measurable Outcomes of an Experiment
 outcome of the experimetnal group
 outcome of the comparison group
* comparison group is not necessarily a "control" group

Measured outcomes depend upon...
the behavior of the participant in response to manipulation of the IV

Sensitivity of DV (2)
must be sensitive enough to detect meaningful differences between groups
best if graded/continous

Ceiling Effect
everyone does well when the DV is not sensitive enough

Floor Effect
everyone does poorly if the DV is too sensitive

Changes in the DV
 found by comparsion
 to different subjects
 to selves in different conditions
The comparsion group determines/drives experimental design.

Different Designs Address Different Problems
 Betweenparticipant (independent groups): different groups of particpants are due to different levels of the IV
 Withinparticipant (repeated measures): a single group of participants is exposed to all levels of the IV.

4 treatments for control groups
 "True" control group: no treatment
 Placebo: give shame treatement
 Waiting list: give no treatment right now
 Yoked: same exact conditions except treatment

Mundane Realism
extent to which your conditions are physically similar to the "real world"
not of primary importance but affects external validity

Experimetnal Realism
extent to which your conditions are psychologically similar to the "real world"
of primary importance: affects internal validity

4 Types of Experimetnal Designs
 Onegroup design: psuedoexperiment
 Onegroup pretest/posttest design: repeated measures/withinpartcipants
 Posttest only with nonequivalent groups: betweenparticipants
 Pretest/Posttest design with nonequivalent groups: measure both, treat one, measure both again; ensures groups are similar to begin with

Comparibility
need to have comparable groups
need to see if effect is larger than or smaller than expected

complete counterbalancing (3)
every possible sequence is used at least once
sequences: (# of conditions) e.g., 3! = 3 x 2 x 1 = 6
 more conditions = more sequences
 e.g., 6! 6 x 5 x 4 x 3 x 2 x 1 = 720

partial counterbalancing
includes only some of the possible sequences
each sequence must appear equally often
use a balanced Latin square to choose sequences

2 Rules for balanced Latin Squares
every condition occurs equally often in each position
every condition precedes/follows other conditions only once

Constructing a Latin Square
 1st row: A, B, L (last), C, L1, D, L2, E, L3, etc.
 2nd row: increase 1 letter at each position of the 1st row (e.g., AB, FA)
 Continue for each row
 Randomly assign each condition to a letter

Graphs are most useful for prediction because...
if the relationship between 2 variables is statistically significant than the score on 1 variable can predict the score of the other

3 Cautions about Interpretations
 don't infer casuality
 beware of directionality problem
 think about 3rd variables

The more sohpisticated the statistical techniques...
...the closer we can come to making casual inferences.

Partial Correlation
 correlatation between 2 variables with influence of 1 or the other variable statistically removed.
 If X and Y are still related after removing Z the correlation is not due to Z.

Oneway/SingleFactor Design
simplest type of design
1 IV, 2 Leveles (experimental, comparison/control)

Comparison Groups (Between v Within)
 Betweenparticipant (independent groups): 1 group receives treatment, 1 does not.
 Withinparticipant (repeated measures): group recieves treatment and also doesn't
 serves as own controls.

4 types of Single Factor Designs
 independent group: betweenparticipant design with random assignment
 repeated mesaures: withinparticipant design
 matched groups: betweenparticipant design without random assignment
 nonequivalent groups: not a "true" experiement

Levels
experimental group + 1 or more comparison group

Increasing levels in a Singlefactor design
 2+ levels give more information
 (e.g., curvilinear relationships: more groups)

Increasing IVs in a Singlefactor design (3)
facorial design: 2+ IV's/factors
closer to realworld conditions
all levels of 1 IV combined with all levels of other IVs

BetweenParticipant Design (Indpendent Groups)
Advantage & Disadvantages (2)
 Advantage
 each participant enters the experiment untainted by previous exposure to the IV.
 Disadvantages
 takes many more participants
 problem of equivalent groups

WithnParticipant Design (Repeated Measures)
Advantages (3) & Disadvantages (2)
 Advantages:
 need fewer particpants
 participatints act as their own controls
 reduces error variance "within" conditions
 Disadvantages
 more demanding to participants
 sequence/order effects

Graphing Correlations (3)
generally use scatterplots
shows relationships between 2 variables
can use only 2 datapoints but this restricts the range and may miss nonlinear relationships

How do graphs improve interpretation? (3)
they help to make informed guesses about whether correlated variables might be casually related
not definitive evidence but can support (or not) a particular casual explination
can help predict hypotheses about possible effects of 3^{rd }variables

Factorial Design Notation
 # of factors noted by # of numbers
 A x B factorial design (e.g., 2 factors A & B)
 # of levels of each factor noted by the #s themselves
 3x2 factorial design
 2 factors: 1 has 3 lvls, 1 has 2 lvls
 combines to 6 conditions (2x3=6)

Betweenparticipants factorial design
 more than 1 IV, all betweenparticipants
 If 1 IV is a particular variable, called a person x environment factorial design

Withinparticipant factorial design
 more than 1 IV, all withinparticipants
 aka repeated measures factorial design

Mixed Factorial Design
 more than 1 IV, both between and within participants
 if 1 IV is a participat variable, it's called a person x environment mixed factorial design

Increasing IV's and Levels
 5 lvls, 3 w/2 factors, 2 w/3 factors
 72 conditions
 720 participants needed for betweenparticipants
 large experiments can become unwieldy
 hard to inerpet: lots of main effects and interactions

Factorial Designs Look At
 main effects: effect of each IV in isolation
 interactions: effect of IV's "combined"

Calculating Marginal Mean
 Marginal mean of B1 (A1B1 + A2B1/2)
 Marginal mean of B2 (A1B2 + A2B2/2)
If the marginal means of B1 and B2 are differnt there may be a main effect.

Calculating Interaction
 E.G.,_A1__A1
 B1___20__40
 B2___30__20
 A1B1=20 A2B1=40 Difference=+20
 A1B2=30 A2B2=20 Difference=10
if the differences found for B1 and B2 are different there may be an interaction.

2 Interaction Rules
If the changes that occur in the means of one row of a factorial table are similar to changes found in the other row(s) the factors are NOT interacting.
If the changes in one row are greater, lesser, or reversed from those in another row there is (i.e., maybe) an interaction

Graphing Interactions
 parallel lines mean no interaction
 nature of relationship for B1 and B2 is the same across levels of A1 and A2

Finding Main Effects & Interactions
 ___________Main Effects____Interactions
 Look at_____marginal means__cell means
 Compare for_sig. differences___patterns of change
 Look at_____SPSS printout__SPSS printout
 Look at _____graphs________nonparallel lines

