1.3 Quantitative Research - Correlational Studies Flashcards
What is a correlation?
A correlation is a measure of linear relationship between two variables and graphically is a straight line that best approximates the scatter plot
How are correlations different from experiments?
Different from experiments as no variable is manipulated
How can correlations be illustrated?
Can be illustrated graphically through scatter plots
What does one coordinate on a correlation graph show?
Coordinate of a spot gives individual person’s score
What is a positive correlation?
Positive correlation - as X increases, Y - increases
E.g if a person gets a high score on one variable they’ll get a high score on the other
What is the main rule when forming a conclusion from correlations?
CORRELATION DOES NOT IMPLY CAUSATION
-only know there’s a link
What is a negative correlation?
when one variable increases the other variable decreases
What is the relation between the gradient and the strength of the correlation?
The steeper the line the stronger the correlation
What range is the correlation coefficient?
A correlation coefficient can vary from -1 to +1
What is a perfect correlation? What correlation coefficient?
A perfect correlation of 1 / -1 is a straight line with the slope of 45 degrees, i.e as one variable increases by one unit, the other variable increases/decreased by exactly one unit
What does the correlation close to 0 look like? What does it show?
A correlation close to 0 is a straight line
Shows that there is no relationship between the two variables i.e if a person scores highly on X this tells us nothing about their score on Y
What are important factors when interpreting a correlation?
- effect size
- statistical signifigance
What is the effect size?
Effect size - absolute value of the correlation coefficient
What is statistical signifigance?
likelihood that a correlation of this size has been obtained by chance
I.e what is chance that if study is repeated with different sample, correlation will turn to 0
What influences statistical signifigance?
sample size
How does a small sample size affect statistical significance?
With small samples cannot be sure that an obtained correlation even if it is relatively large, has not been obtained due to random chance
How can we determine the probability of a correlation being obtained by chance?
The probability that a correlation has been obtained due to random chance can be estimated, but there are conventional cut-off points when results are considered to be “statistically significant” or not
Convention cut-off point for statistical significance is 5%
What types of correlations do scientists look for?
Scientists look for statistically significant correlations with large effect sizes
What are 3 limitations of correlational studies?
- third variable problem
- curvilinear relationships
- spurious correlations
What is a third-variable problem?
a third variable may correlate X and Y and explain the correlation between them
What is a curvilinear relationship?
variables are sometimes linked non-linearly and if a straight line is drawn, we would obtain a small to medium correlation coefficient
What are spurious correlations?
When a study involves calculating multiple correlations between multiple variables, there is a possibility that some of the statistically significant correlations would be the result of random chance
When do spurious correlations occur?
If we only choose correlations that fit our hypothesis (bias)
When we calculate 100
correlations and only pick ones that turned out to be significant, increases the chance that we have picked spurious correlations
Methodological concern
How is sampling generalisability decided in correlational studies?
Firstly target population is identified depending on the aims of the study and then a sample is drawn from the population using a sampling type (same as in an experiment)
What is generalisability in correlational studies linked to?
Generalisability of findings in correlational research is DIRECTLY linked with sampling and depends on REPRESENTATIVENESS of sample (population validity)
On what levels can bias in correlational research occur?
Can occur on the level of variable measurements (bias in observational data) and on the level of interpretation of findings
What 3 biases are on the level of interpretation of findings?
- curvilinear relationships
- third variable problem
- spurious correlations
How can curvilinear relationship (between variables) bias be reduced?
If suspected researchers should generate and study scatter plots
How can a correlational study be more credible (third variable problem)?
research more credible if researcher considers potential third variable in advance and includes them in research in order to explicitly study links between X and Y and third variable
How can spurious correlation bias be reduced and credibility increased?
To increase credibility, the results of multiple comparisons should be interpreted with caution
Effect sizes considered together with level of statistical significance
