Chapter 9 🤎

Why can't a simple bivariate correlational study meet all three criteria for establishing causation?

(1) Covariance: yes, proved that there is a correlation! (2) Temporal Precedence: cannot say which variable came first and caused the second variable (3) Internal Validity: third variables are not usually controlled for!

Mediator

** A variable that helps explain the relationship between two other variables (as a go-between for two variables) - explains the process through which two variables are related - Can be detected using multiple regression -Mediators are internal to causal variable Ex: sleep quality (an independent variable) can affect academic achievement (a dependent variable) through the mediator of alertness. In a mediation relationship, you can draw an arrow from an independent variable to a mediator and then from the mediator to the dependent variable.

Autocorrelations

** To see if things change overtime - correlations of the same variable at different points Ex: the temperatures on different days in a month are autocorrelated

Cross Sectional Correlations

** cannot establish temporal precedence -Looking for correlations between different variables at the same time Ex: looking at both communication frequency and relationship satisfaction at time one

Cross-lag correlations

**Closer to establishing temporal precedence - In Longitudinal design they show whether the earlier measure of one variable is associated with the later measure of the other variable. Ex: whether watching televised violence causes aggression or aggression causes people to prefer viewing television violence.

Longitudinal design

**provides evidence for temporal precedence - A study in which the same variables are measured in the same people at different points in time. Ex: a five-year study of children learning to read would be a longitudinal study. Researchers might compare environmental and other factors in the children and measure outcomes over time.

Ruling out third variables with (Multiple Regression) (2)

- "Control for" → try to see if the relationship still exists when something is held constant Ex: Third variable: age: sexual content and pregnancy are correlated bc teens are more sexually active. Sex and age are correlated as well as pregnancy and age. Researchers want to find out if the third party: age is correlated with the original variables: sexual tv content and pregnancy rates. They see what happens when they control for age

Multivariate Designs

- (such as longitudinal and multiple regression designs) involved more than two measure variables Ex: we cannot predict the weather of any year based on the season. There are multiple factors like pollution, humidity, precipitation,

Claim 3 " Amount of TV people watched in their young adult years was linked to lower cognitive abilities in middle age"

- Bivariate (associations that involve exactly two variables) - Negative association between variables - Criterion Variables → score on cognitive tests - Predictor Variables → average number of hour per day watching TV (and all of the variables that were controlled for)

Example Claim "Active sex life may lead to improved job satisfaction and engagment in work"

- Criterion Variable → job satisfaction and engagement in work - Predictor variable → active sex life - The study is longitudinal (data was collected multiple times from the same groups) - Mediator → positive mood

2. Temporal Percedence

- If A caused B then A must come before B - The cause occurs before the effect - Cause and Effect are related and/or covary Ex: What happened first, the chicken or the egg?

Regression does not establish causation

- Multiple regression is not a fool proof way to rule our all third variables - Regression tables control for the variables mentioned, but there is no guarantee that there are other variables that were not controlled for involved

Beta basic

- Tells us the direction and strength of a relationship (the stronger the beta the stronger relationship between criterion and predictor variable) - No specific guidelines like there are with r - Can only compare beta strengths within a single regression table

What are two reasons that multiple regression analysis cannot completeley establish causation?

- They cant establish temporal precedence; researchers cant control for variables they don't measure (there could be a third variable that they didn't measure that is responsible for the relationship)

Getting a casualty with pattern and parsimony

- We can be more sure about causal claims if there are multiple studies coming to the same or similar conclusions Ex: Journalists do not always flirt represent pattern and parsimony (When journalists report only one study at a time, they are selectively presenting only one part of the scientific process (only use things that are entertaining for the audience)

Multiple Regression

- helps rule out some 3rd variables and address questions of internal validity

Parsimony

- simplest explanation of some phenomeon - casual claim= simplest explanation of a pattern of data- theory that requires making the fewest exceptions or qualifications

3. Internal Validity

- the degree of confidence that the causal relationship being tested is trustworthy and not influenced by other factors or variables **Must be able to control for confounding variables Ex:You want to test the hypothesis that drinking a cup of coffee improves memory

1. Covariance

- when two factors have a relationship to each other and one changes, there should be a change seen in the other factor also, either positive or negative. Ex: a study shows that a supportive adult figure has a positive relationship or positive Covariance with a child's having good grades in school. Positive Covariance can be shown because when the adult is more supportive, grades go up; when the adult is less supportive, grades go down

Ruling out third variables with (Multiple Regression)

-When you have more than two variables and can control the confounds (outside influence changing effect) , you can get internal validity (trustworthy cause-and-effect relationship between a treatment and an outcome.) Ex: an educational psychology researcher could use multiple regression to predict college achievement (e.g., grade point average) from the variables of high school grade point average, Scholastic Assessment Test (SAT)

How many Criterion variables are there in a multiple regression analysis? How many Predictor variables?

1 criterion variable and 2 predictor variables

Give at least two phrases indicating that a study used multiple regression analysis?

1) "Controlled for" 2) "Taking into account" 3) "Correcting for"

Regression results indicate if a third variable affects the relationship (Multiple Regression)

1. Criterion variables (dependent v)- ex: predicting pregnancy 2. Predictor variable (independent v) - ex: amount of sexual content each teen reported watching on TV and the age (variable) of each teen - Is pregnancy affected by the amount of watched sexual content and age? - Use beta to test for third variables (confounds that could be related to both variables)

what are the three kinds of correlations obtained from longitudial design? what does each correlation represent?

1. Cross-lag correlations 2. Auto correlations 3. Cross-sectional correlations

Why not just do an experiment?

1. In many cases participants cannot be randomly assigned to a variable Ex: While parents might be able to learn new ways to praise there children, they cant easily be assigned to daily parenting styles, so its hard to manipulate this variable. 2.Cannot be assigned to preferences Ex: Unehtical to assign children to a condition where they recieve a certain type of praise especially over. along period of time especially if we suspect that type of praise will make a child narcasitic. 3.Unethical to assign participants Ex: If researchers suspect that smoking causes lung cancer or sexual content on TV causes pregnancy. It would be unethical to ask study participants to smoke cigarettes or watch certain TV shoes for several years.

Multivariate Designs and the Four Validities

1. Internal Validity → with longitudinal design we can get closer to making a causal claim 2. Construct Validity → ask whether the measures that are used are reliable 3. External Validity → we want to know how the sample was collected/ can the findings be generalizable 4. Statistical Validity → are the results significant

Variables

Communication frequency and relationship satisfaction Data collection time points → Jan, Feb, March and April

3. Give an example of a question you would ask to interogate each of the four validities for a multivariate study?

Construct validity: how well were the variables measured? External validity: to what degree do the findings generalize? Statistical validity: what was the effect size? was the effect size significant? Internal Validity: does it establish temporal precedence? does it rule out any third variable?

1. State why simple bivariate correlations are not sufficient for establishing causation

Correlation does not imply causation. Cannot determine temporal precedence

1. Think of a possible mediator for the relationship between exposure to sex on TV and the chance of pregnancy. Sketch a diagram of the mediator you propose.

Diagram should resemble figure 9.12 with exposure to sex on tv in the left box, pregnancy risk on the right box and your proposed mediator in the middle box.

Describe which patterns of temporal precedence are indicated by different cross-lag correlational results

Either variable 1 at time 1 is correlated to variable 2 at time 2, or variable 2 at time 1 is correlated to variable 1 at time 2, or there might be correlations between both variables at both times when cross lagged with each other (i.e., all of the above mentioned).

What is a responsible way for journalists to cover single studies on a specific topic?

Ideally, journalists should report on the entire body of evidence, as well as the theoretical background, for a particular claim

Why do many researchers find pattern and parsimony an effective way to support a causal claim?

In psychology, researchers commonly use a variety of methods and many studies to explore the strength and limits of a particular research question. Often using a broader, simple explanation for how variables relate can keep confounds/third variables limited that occur when one gets more specific with a question.

Predictor variable

Independent variable; variables measured in a regression analysis - variables that are being used to predict some other variable or outcome Ex: predictors such as qualifications, relevant work experience, and job-specific skills (e.g., computer proficiency, ability to speak a particular language) may be used to estimate an applicant's future job performance.

Describe what it means to say that some variable "was controlled for" in a multivariate study?

It means a variable the researcher though of and pays attention to, that may have an effect on the outcome of the results of the study.

Why is a longitudinal design considered a multivariate design?

Longitudinal designs are multivariate because they even if they measure the minimum two variables to check for correlation, they measure them at two different points in time, which actually means they are measuring four variables all together.

2. Explain how longitudinal correlational designs help address temporal precedence

Measure the same variables at different points in time

2. What three variables would you have to measure in order to test your mediator hypothesis, and when would you have to measure then?

Three variables should be measured at different points in time: first sex on TV, followed by your mediator, and then chance of pregnancy

What is the relationship between 95% CI for Beta and the betas statistical significance?

a range of values that you can be 95% confident contains the true mean of the population

Three Causal Criteria

covariance, temporal precedence, internal validity

Criterion Variable

dependent variable; the variable they are most interested in understanding or predicting Ex: we may use the predictor variables hours studied and hours of sleep the night before the test to predict the value of the criterion variable test score. In this case, our criterion variable is the variable being predicted in this analysis

5. Explain the function of a mediating variable

explains the process through which two variables are related

3. Explain how multiple-regression analyses help address internal validity (the third-variable problem).

multiple regressions can help rule out 3rd variables; measuring at least 3 variables (one is dependent variable, others are independent or predictor variables