Research Methods Exam 3

Front 1

Factorial design

Back 1

Has more than one independent variable

Front 2

Notation of Factorial Design

Back 2

- e.g., 2 X 4 has 2 independent variables, one with 2 levels, one with 4 levels
so a 2X4X6 – has 3 I.V.s – 2 levels (eg. Male vs. female), 4 levels (eg. One of 4 possible dosses) and 6 levels (one of 6 behaviors tasks difficulty levels) – 48 conditions

Front 3

Main Effect

Back 3

overall effect of one independent variable

- To find if there is a main effect in the rows you look at the individual row means and if there is a difference then there is a main effect.
- To find if there is a main effect in the columns you look at the individual column means and if there is a difference then there is a main effect.

Front 4

Factorial experiment with all between-subjects independent variables (I.V).

Back 4

- Looking at effect of attention-enhancing drug (e.g. “focusol” on cognitive performance). You hypothesize it will help people focus in distracting environments, but won’t have any effect in non-distracting environments.
- First I.V. = Drug (given vs. not given)
- Second I.V. = distractibility of environment (high vs. low)
- In a 2X2 between-subjects factorial design (where both I.V. are between subject variables), each subject only experiences one of the combinations of levels (ie only one of the conditions of the experiment, or only one of the cells in the factorial matrix.)

Front 5

Repeated measures factorial design

Back 5

If all I.V.s are within-subjects variables

Example:
-First I.V. = Drug (given vs. not given)
-Second I.V. = distractibility of environment (high vs. low)
- Day 1: give drug and stroop test in each type of environment
- Day 2: give placebo and stroop test in each type of environment
-Each subject experience each combination of levels or condition (ie each in the matrix)
- Have to counterbalance conditions appropriately – this example would be called a
2X2 repeated measures factorial.

Front 6

Mixed factorial design

Back 6

If at least one I.V. is between-subject and at least one is within-subjects


Example:
-First I.V. = drug as a between-subjects variable
-Second IV = environment, as a within-subject variable
-Half the subjects get drug and complete stroop task in each of the environments; half the
subjects get no drug and complete stroop task in each environment.
- ie, each subject experiences 2 of the 4 possible combinations of levels of I.V.s (i.e 2 of the 4 cells in the matrix) – this example would be called 2X2 mixed factorial

Front 7

Positive correlations

Back 7

As one variable increases, so does the other.

E.G. foot size is positively correlated with height, E.G. self-esteem is directly related to “life satisfaction”

Front 8

Negative correlations

Back 8

As one variable goes up, the other goes down

- can also say the variable are inversely related, or indirectly related. E.G. measures of self-esteem negatively correlate with measures of depression.

Front 9

No correlation

Back 9

–increases (or decreases) in one variable are unrelated to changes in other variable. E.G. there is no correlation between foot size and I.Q.

Front 10

Pearson’s r is used to measure strength of correlation

Back 10

- r ranges form -1.0 to 1.0 and can be any point in-between.
-positive correlations are represented by positive r-values (eg. 0.36) and negative correlations are represented by negative r-values (e.g. -0.83)

Front 11

Know how to interpret r (goes from -1 to 1)

Back 11

-Larger values (whether positive of negative) reflect stronger associations
->0 to 0.5 = weak positive correlation
->0.5 to 0.8 = moderate positive correlation
->0.8 to 1.0= strong positive cor.
->1.0 = perfect pos. cor. Even -1.0 is perfect

Front 12

Coefficient of determination (r2)

Back 12

Is obtained by squaring r.

- means that some % of variability in one variable can be accounted for, or predicted from, the variability in the other variable.
- EG 0.65*0.65 = 0.42 – 42% of variability in GPA is related to variability in SAT score

Front 13

Why can’t we conclude causation from correlation?

Back 13

To infer that one variable (eg lonliness) causes another variable (eg depression), you need to show:
– 1. that the two variables covary: if X goes up, Y goes up also (or down if it’s an inverse relation)
• Correlational studies can do this
• Loneliness and depression correlation r = 0.6
– 2. Directionality: X precedes Y
• Correlational studies can partly, “sort of” do this
• “directionality problem”
• Not sure if loneliness came first, then depression, or vice-versa
– 3. All other extraneous variables can be ruled out
• Correlational studies can’t do this
• “Third variable problem”
• Perhaps some third factor (e.g., a brain injury) caused both depression and loneliness

Front 14

Combined interrupted time series with non-equivalent control group design

Back 14

Improves confidence that it’s something about D.C., but can’t positively rule in the click-it-or-ticket campaign
• Could still be differential history effect.
E.g., coincidentally, DC decided to start cracking down on all traffic violations at same time campaign started

Front 15

Pros of small N research

Back 15

1) Cheaper (subjects cost money)
2) Sometimes can only find one or a few subjects with certain characteristic
• E.g., Phineas Gage- mellow before accident, after not so much
3) Individual subject validity not a problem

Front 16

AB design

Back 16

Simplest design, just 2 phases, baseline and then the treatment.
• EX measure baseline rate of disruptive behavior in classroom over several days, then implement “Differential-Reinforcement-of-Alternative Behavior” treatment and measure behavior for several more days.

Front 17

Changing criterion design

Back 17

Somewhat analogous to “shaping” or “successive approximation”
Behavioral criterion is successively changed over phases.
• Provide reinforcement ($5) for progressive reductions in smoking (baby steps- once reach certain level, stop)

Front 18

Naturalistic observation vs. participant observation

Back 18

Naturalistic Observation- Observing people or animals in their natural habitat.
• E.g., Jane Goodall and chimpanzee observations (tool use)
Participant Observation- Observer becomes part of group and records experiences.
• E.g., Cult infiltration
• Famous David Rosenhan experiment with psychiatric hospital.
o Had people admitted into psych ward and found how once put in ward, hard to get out of system

Front 19

Participant (Subject) Reactivity

Back 19

Subjects may behave differently knowing that they are being observed.
• Humans and animals.
Solutions:
• A) Hide
o Unobtrusive measures, e.g., hidden camera, two-way mirror
o Ethical issues
• B) Habituation
o Subject is aware of, but used to observation and acts naturally.

Front 20

Observer bias

Back 20

Many behavioral observations involve some subjectivity.
• Methods for dealing with observer bias
o Establishing strict operational definitions of construct
o Behavior checklists
o Checking inter-observer (rater) reliability
o Automation of observation

Front 21

Simple random sampling

Back 21

• Each individual in population has same probability of being included in sample.
• Need to know exact size of population and have list of all its members.
• Still possible to end up with unrepresentative (biased) sample by chance, especially if small sample

Front 22

Stratified sampling

Back 22

A way to help ensure that a random sample is not biased with respect to some important population characteristic.
• E.g., female/male ratio.
• If AU is 60% female, then a representative sample of AU students should have 60% females.
• If sample is small, could end up with unrepresentative sample by chance (e.g., 70% male, 30% female).
o First, determine desired total sample size (100),
o Then figure out how many females you need (60), how many males you need (40) to ensure a sample that is representative with respect to gender.
o Then randomly select females until you have 60 females and randomly select males until you have 40 males.

Front 23

Convenience sampling

Back 23

• Weak form of sampling, but probably most commonly used in psychology.
• Use those subjects that are easy to get
o E.g., students taking psychology classes.
• Easy to end up with biased sample.

Front 24

Quota sampling

Back 24

Like the stratified sampling procedure, but done with a convenience sample.
• E.g., want to ensure that for a total sample size of 30 that 60% of subjects female and 40% are male,
• Stand outside library and survey females until you get 18 and survey males until you get 12.