Correlational Research Flashcards
(34 cards)
What is correlational research and how does it differ from experimental and descriptive?
descriptive: “what”
correlational: “how”
experimental: “why”
Correlational:
goal is to examine and describe the associations and relationships between variables. It doesn’t attempt to explain the relationship, and make no attempt to manipulate, control or interfere with the variables (like experimental research would)
- usually two scores per participant (in two variables) if three, then three scores and so on….
if you see in a question “randomly assigned to two groups”: clue that it is experimental, as with correlational, no manipulation.
Why do we bother looking for correlation between variables?
To assess difference between correlation and causation.
phrenology - was said to be causation. but there is no such thing as a particular part of the brain translating into a specific behavior. Maybe there is a correlation, but causation……. Causation can only be determined using experimental research (manipulating). So beware of words like: “it is the cause of….” If you don’t have appropriate data to support it.
How does a correlational research work?
Measure number x different variables for each individual. The researcher then looks for a relationship within the set of scores
What is the difference between correlation design and differential design?
Correlational views the data as two scores (or more depending on how many variables used) , for each individual, and looks for patterns within the scores to determine whether there is a relationship.
Differential design uses one of the two variables to define groups of participants and then measures the second variable to obtain scores within each group.
How does data of a correlational study look like?
Each participant has a score for each variable.
What is a scatter plot?
It is a representation (a way to visualize) the intersection of two variables, x and y (e.g. brain weight and IQ score). Essentially, each plot represents one individual.
What is the line of best fit?
Line that shows where/what the overall trend is. The idea is the closer the dots are to the line, the stronger the association!
What are three possible directions correlation can take? Describe each one!
- Negative correlation: A decrease in one variable causes an increase in another.
- No correlation (proche de 0)
- Positive correlation: an increase in one variable is linked to an increase in another variable or one decreases and so does the other.
What does “r” refer to?
r = correlation coefficient that tells us about observed relationship
Possible range: -1 to 1
- : indicates a negative association/ correlation. Une pente négative en gros indicates variables move in opposite directions.
In regards to the strength of correlation, what is the consensus for a moderate to strong association?
0.8 and more
Perfect negative relationship : -0.8 to -1
Medium negative correlation
No correlation : 0 (pas de pente, flat)
Medium positive correlation
Perfect positive correlation : 0.8 to 1
Is it possible to have a correlational coefficient of 1?
Well apart si tu compare la variable avec elle meme non. Cause there is always some kind of error that affects the outcome.
What is the correlational coefficient used when relationship is not linear?
Spearmen correlation coefficient (s)
Pearson correlation coefficient only works for linear relationships, but not all relationships are linear!!!
Spearmen correlation coefficient also ranges from -1 to 1.
What is a monotonic relationship? What is the link with ordinal scale.
In a monotonic relationship, the variables tend to move in the same relative direction, but not necessarily at a constant rate
Ordinal scales are about ranking (1st, 2nd, 3rd), and monotonic relationships care about the order of changes (more or less).
Example:
If we ask kids to rank their favorite toys (ordinal scale), we can check if liking a toy more means they play with it more (a monotonic relationship).
Even if we don’t know exactly how much more they like it, we can see the trend.
Ordinal scale = Ranks (like 1st, 2nd, 3rd).
Monotonic relationship = A pattern where more of one thing means more (or less) of another, consistently.
They connect because both focus on the order of things, not exact numbers
Correlation tells us the degree to which…
two variables are associated
What is the percentage of shared variance and how is it calculated?
(R²)
example:
r= 0.8
(R²) = … x 100 = 64% (tells us what to what extent the variables tend to vary together)
Diff shared variance and coefficient of determination.
shared variance: %
coefficient of determination: degree of overlap between variables
if you see r=… signals what???
That it is a linear relationship.
What is the general trend for r and r2
As r increases, r2 increases as well.
Does a shared variance of a 100% mean that what is observed is causation?
NO. still only an association. Prcq ok lets say the correlation between number of friends and IQ is perfect. Well you could have a high I1 because you have many friends, or you have many friends because you have a high IQ. So which one is it? Correlation design doesn’t allow you to conclude which one causes the other it only tells you that both are perfectly related.
What is statistical significance?
The result is very unlikely to be produced by random chance.
Does strong correlation, necessarily mean strong statistical significance?
No, not necessarily. It all the depends on the alpha level. Alpha level refers to the maximum probability that the result is produced by chance that is accepted (false positive). Usually 0.05. coefficient r must be larger than the critical value in the table. If that is the case then good, we reject the null
p-value smaller than a, so we reject the null
What is a spurious correlation?
The false assumption that there is an association when there isn’t, because of a third variable, coincidence, or data manipulation. Even if there is statistical significance, doesn’t necessarily mean that it isn’t a spurious correlation. Because ok lets say ice cream sales and shark attacks spurious correlation because both are influenced by summer weather, yet still statistical significance . C correct cause statistical significance only tells us about the probability that the results aren’t due to random chance.
bref, highlights the point that just observing a correlation is not enough to conclude a causal effect (you cant say ice cream causes polio just because you observed a [spurious] correlation) .
Is having a strong correlation and statistical significance enough to conclude causation?
No!!!!!!!!!!!!!!
What are the two problems that don’t allow for us to conclude causality even with strong correlation and statistical significance?
Directionality problem
Third variable problem
