PS125 Research & Statistical Methods Term 2 Part 1

Question 1

Experiments

Answer 1

Experimenter manipulates one (or more) independent variables
We measure onr or more variables
Look for difference between groups (between) or conditions (within)

Question 2

Quasi-experiment

Answer 2

Experimenter uses one (or more) pre-existing variables (IVs)
We measure one or more variable’s (DV/s)
Look for a difference between groups (between) or conditions (within)

Question 3

Correlational studies

Answer 3

Experimenter measures two or more variables
No manipulation, no separate groups
Variables are continuous, not categorical
Look for a relationship between co-variables

Question 4

Why used correlational designs?

Answer 4

Continuous data
Can’t always manipulate variables
Allow us to make predictions - using graphs y = mx + c to plot in data of one variable to find the other variable by using scatter graph
Nuanced relationships - not necessarily cause and effect

Question 5

Correlation limitations

Answer 5

Correlation does not always equal causation
Direction problem - is the first variable causing a change in the second variable or the other way round?
Third variable problem - is there a different variable that we haven’t measure that is causing this correlation
Spurious correlations - correlations can be found between almost anything e.g popularity of the first name Andrea and cottage cheese consumption

Question 6

How do we know if correlation is a causation?

Answer 6

Time precedence - one thing has to precede the other e.g how much revision done and exam results
Statistically significant relationship
Non-spurious

Question 7

Questionnaire Design

Answer 7

Not straight forward!
Needs to:
Be valid
Be reliable
Capture variability
Be the right length
Avoid leading questions
Avoid confusing or ambiguous questions

Question 8

How to design a questionnaire?

Answer 8

Design one yourself
Adapt other measures
Check for clarity
Collect pilot data
Check distribution of individual questions
Prinicipal Component Analysis to form subscales
Confirm external validity
Confirm internal reliability
OR
Use an existing measure

Question 9

Interpreting a correlation

Answer 9

Is there a statistically significant relationship between variables? If yes:
What is the direction of the relationship?
What is the strength of the relationship? -1-1 range -1 negative correlation and 1 positive correlation
What does it mean?
NB: Correlation is NOT Causation

Question 10

Scatter plots/graphs

Answer 10

Sumarise bi-variate data, by plotting individual data points in a two-dimensional space

Question 11

What are variables referred to in a scatter graph?

Answer 11

Notice that variables are referred to as ‘predictor’ and ‘criterion/outcome’

Question 12

Positive vs negative correlation

Answer 12

Positive relationships - as X increases, Y also increases
Negative relationships - an increase in X is generally associated with a decrease in Y

Question 13

Perfect relationships

Answer 13

Perfect relationships
Perfect positive - increase in X is accompanied by exactly proportional increase in Y
Perfect negative - decrease in X is accompanied by exactly proportional in crease in Y

Question 14

No relationship

Answer 14

No Relationship: no relationship between X and Y - no systematic tendency for Y to vary with X

Question 15

Correlation

Answer 15

The relationship between variables (two variables: bi-variate)
Correlation coefficient = numerical measure of the direction and degree of strength of the relationship
Many different correlation coefficients exist
Most used is Pearson product-moment correlation coefficient (symbol: r)
Range of r: -1 to 1

Question 16

Covariance

Answer 16

Covariance is related to the concept of variance - the average amount that data can vary from the mean

Question 17

Testing the significance of a correlation coefficient

Answer 17

Correlation of a sample r, is unlikely to be identical to the population correlation p (rho) due to sampling area
Common research question: are two variables relationed in population
Null hypothesis H0 p = 0

Question 18

Factors that affect correlations

Answer 18

Range restrictions
Non-linearity
Correlation is only an index of linear relationships between variables (in raw data or ranks)
Sometimes variables show curvilinear relationships -n in those cases although relationship can be strong and systematic, correlation can be small
Heterogeneous subsamples
Correlation for the entire data set can differ from correlations within each subset
Example: correlation between height and weight is .76 bt r is lower with male and female subgroups
Outliers

Question 19

Correlation - limitations

Answer 19

Correlation does NOT always equal causation
Direction problem
Third variable problem
Spurious correlations

Question 20

Partial correlation

Answer 20

“Third variable” problem
Spurious correlation e.g drowning, ice cream sales and temperature
Gets rid of a variable which better explains co-variance
Obscured correlations e.g age, height and diet gets rid of a nuisance variable which obscures co-variance