Chapter 8: Bivariate Correlational Research Flashcards
1
Q
What is the difference between experimental methods and correlation?
A
- Experimental methods test if changes to IV causes changes to DV score. Ex. Does coffee make you anxious?
- Correlations are descriptive methods. Look at behaviors as they occur. Ex. Is there a relationship between feeling anxiety (X) and coffee drinking (Y)?
2
Q
What do correlations do?
A
- Evalulate the strength of a relationship
- The mathematical measure of an association between 2 variables
- Do NOT establish cause
3
Q
When would you use a correlation?
A
- To make a prediction
- For validity: If you made a new IQ test, to determine how good it is, you could compare results on the new test (X) to results on another valid test (Y)
- For reliability: Measurements are stable if they produce similar results across situations. Ex. To see if your score today is correlated with your score next week
4
Q
What are some association claims supported by correlational research? Why aren’t they causal statements?
A
- Meaningful conversations linked to happier people
- Couples who meet online have better marriages.
- Tehse are not causal statements because there was no random assignment to different groups
5
Q
Bivariate correlation/association
A
- Three kinds: pos, neg, zero
- 2 variables (X and Y). Each participant will have a score on both variables. Asks: Is there a relationship between the way one variable changes and the way a second variable changes?
6
Q
What does a scatterplot do?
A
Represents the score for each participant on both variables
7
Q
Cacioppo et al, 2013
A
- Found that ouples who meet online tend to have between marriages.
- Emailed surveys asking “How did you meet?” and the Couples Satisfaction Index (CSI)
- 19,000 respondents… people who met online tended to score slightly higher on CSI
8
Q
What two pieces of information does the corrleation coefficient tell you?
A
- Strength of the correlation
- Direction of the correlation
Is a value between -1 and 1
9
Q
How does the correlation coefficient (r) tell you the strenght of the correlation?
A
- Closer to +/-1.0= strong relationship
- Closer to +/- 0.01= weak relationship
10
Q
How does the correlation coefficient (r) tell you the direction of the correlation?
A
- Positive: values increase and decrease together
- Negative: one value increases, the other decreases
11
Q
How to visualize associations with categorical data?
A
- Ex: Did you meet online? Categories are yes and no.
- Scatter plots can still be useful, as you can seee a higher density of dots in online.
- But bar graphs are usually more helpful for categorical variables
12
Q
Do correlational studies establish internal validity?
A
NO
13
Q
What is construct validity?
A
- the degree to which a test or measurement tool accurately assesses the theoretical concept or “construct” it is intended to measure
14
Q
How can you determine construct validity of association claims?
A
- Because association claims describe the relationship between 2 variables, it is reasonable to ask about the construct validity of each variable.
- Can ask: How well were each of the variables measured? Does the measure have good reliability? Is it measuring what it intends to measure? Face, discriminant, and convergent validity?
15
Q
Using the CSI example, what questions should you ask to determine its’ construct validity?
A
- Does it have good internal reliability?
- Does it correlate with other measures of marital happiness?
16
Q
What is face validity?
A
- Compared to construct validity, which examines whether the tool truly captures the underlying theoretical construct, face validity is a more superficial assessment based on how the tool appears on the surface
- Face validity is a subjective evaulation of whether a test “looks like” it measures what it intends to
17
Q
What is discriminant validity?
A
- The extent to which a test or measurement tool accurately differentiates between distinct constructs
- Meaning, it should now show high correlations with tests designed to measure different, unrelated concepts
- Demonstrates that a test is measuring what it intends to measure and not picking up on other unintended constructs
18
Q
What is convergent validity?
A
- Measure of how closely a test or instrument correlates with other tests that measure the same or similar constructs
- Established when there is a strong positive correlation between the results of a test and other tests that measure the same construct
19
Q
Questions to ask to determine statistical validity
A
- How well do the data support the conclusion?
- How strong is the relationship?
- Effect size (the strength of a relationship between two or more variables)?
20
Q
Statistical validity and effect size
A
- All else being equal, a larger effect size is better/more important
- But a small effect size over multiple observations can also be important (Ex. taking 0.5 sceonds off an olympic athlete’s sprint time)
21
Q
Statistical Validity: What is aggregation and what are some examples
A
- Aggregation: small effect sizes can compound over many observations
- Ex. Baseball: MLB players have higher probability (r=0.05 of getting to a base on any given turn at bat compared to minor league baseball. This is a huge difference over 550 turns at bat.
- Ex. Agreeableness: Agreeableness= better social interactions (r=0.07) in mid size uni first semester. But there are 100s of interactions in the first few weeks, so that is many ore positive interactions
22
Q
Aggregation: Many participants
A
- Teens in US high schools with a growth mindset (can be smarter through effort and good learning strategies) have better grades (r=0.05)
- If implemented at scale, this could be enough to prevent 79,000 teens from scoring in F and D range
23
Q
Statistical validity: How precise is the estimate?
A
- How right/wrong could your estimate possible be?
- 95% confidence interval= margin of error (range) of the estimate. Included true values that will be obtained 95% of the time
24
Q
Confidence Interval Example
A
- Time spent sitting and medial temporal lobe thickness
- R=-0.37 [CI=-0.07, -0.64]
- Notice range does not include zero. When it does contain zero, it is possible there is no relationship
really big range bc data only represent 35 ppl
25
Statistical Validity: Sample size and precision
* When samples are small it is less stable
* Outliers have a bigger CI
* Smaller samples have bigger CIs
* Bigger samples have smaller and more precise CIs
26
Statistical validity: replication
* Important to see if the study has been replicated.
* Ex (Milek et al 2018) found there is a relationship between deep talk and overall wellbeing.
* Ran repeatedly... 4/5 reflect a positive relationship
27
Statistical Validity: outliers
* Can the results be due to an extreme score?
* Depending on the variable and the number of observations, a single score can skew the results
* In association claims, outliers have a bigger impact when they are extreme on both variables (ex someone who is v tall and v heavy)
* Outliers matter most when the sample is small
28
Statistical Validity: Range of data
* If there is not a fully represented range on one of the variables it can make the effect seem smaller
* Can apply at any time one of the variables has little variance
* SAT scores are predictive of performance in 1st year (r=0.33) but also only includes ppl w/ scores >1200. When you expand the range the correlation gets stronger
29
Can we make causal inference from associations/What are the three criteria for causation?
1. Covariance of cause and effect
2. Temporal precendence- have a directionality problem
3. Internal validity- problem w third variable
30
Why does correlation not equal causation?
* Correlations are descriptive, not experimental
* Often there is another variable that is influencing X and Y
31
Correlation does not equal causation example
* Are disney princess movies related to agreement with female empowerment?
* 3rd variable; boys who watch these likely have sisters
32
What type of test is correlation?
* Descriptive, set up to describe data
* Not set up to ask "why" questions
* Some very stange things are closely correlated
33
Weird correlations
* Ppl who drowned after falling out of a fishing boat and marriage rate in kentucky
* Searches for JK rowling and coca cola stock price
34
What is evaluative conditioning?
Pairing products with another thing people like
35
Why would someone think correlation = causation?
* This tendency reflects a bias of brains
* The brain is a pattern seeker, and likes finding causal relationships even when none exist
* The brain is neurologically tuned to notice when things occur together (Hebb Rule: neurons that fire together wire together)
* We use statistics to overcome our biases
36
Internal validity w association claims: the third variable problem
* Height and hair length= negative correlation (but due to 3rd variable- gender)
* But 3rd variable is not always a problem (height and weight- despite gender, generally taller women weigh more and taller men also weigh more)
37
What is a spurious association?
The association is there but it is not "real"
38
Association claims and external validity
* Can an association be generalized? To other places, ppl, and times?
* Size of sample matters less than how participants were selected. Random assignment is best
39
What is a moderating variable?
* When the level of a variable changes because of the level of another variable
* A 3rd variable that influences strength or direction of the relationship between X and Y
* Suggest that the relationship between 2 variables changes bc of the moderator
40
Moderating variable: fandom example
* The better the team does, the more tickets sold
* But, residential mobility matters, as it increases fairweather fans
* Moderating variable= residential mobility
