Multivariate Correlational Research

Chapter 9 part 2: multivariate correlational research In class activity #5 The amount of TV people watched was linked to lower cognitive abilities. Proposed third variables is alcohol use  1) what is the criterion DV and what is the independent IV o cognitive abilities; alcohol use and amount of tv  2)which pattern of results would indicate that only amount of TV people watched independently predicts cognitive abilities, controlling for alcohol? o  only the beta for amount of TV is significant 3) )which pattern of results would indicate that only alcohol use independently predicts cognitive abilities controlling for amount of tv people watched? o only the beta for alcohol use is significant  4) both betas are significant Regression does not establish causation  Review: are multiple regression studies able to show causation? o Temporal precedence? maybe not o Internal validity: you can only control for variables that you thought to measure. Only variables thought to measure. Underlined variable clicker  Conscientious people live longer because they engage in healthy behavior. The underlined variable addresses the “why” of a relationship is called? o Mediator Introduction to mediation Mediation = WHY  Start with an association between two variables, A and B  Mediation hypotheses propose a mechanism for a bivariate relationship. Why are these two variables correlated? o Often use tools as regression to analyze mediators.  Mediation hypotheses are causal statements o Arrows in mediation diagram only go in one direction.  Mediators specify a time sequence for the three variables (temporal precedence)  Mediators also specify the mechanism A mediator vs third variable   Physical activity as a mediator  Economic disadvantage as a third variable o THIRD VARIABLE= EXTERNAL TO THE TWO VARIABLES IN THE BIVARIATE RELATIONSHIP. Recess and behavior are only correlated because they are both separately associated with income. o Related to both of them. Relationship is External!  Mediator is internal to the causal variable, and usually of direct interest to the researchers. Mediator vs moderator  Moderator: is the extra variable that depending on its level is going to change the relationship of the two variables of interest. Looks for differences in particular subgroups  Mediator: asks WHY the variables are associated  Physical activity as a mediator  Marital status as a moderator Mediator, third variable or moderator?  Viewing violent television is associated with aggressive behavior because children model what they see on tv. o What is your Simple bivariate correlation (variables)?  Viewing violet tv, and aggressive behavior o What’s the extra variable... and what is it?  Mediator   A is related to B because … Viewing violent television is associated with aggressive behavior because people who watch more television have more lenient parents, and these parents don’t care if their children are violent. o What is your Simple bivariate correlation (variables)?  Viewing violet tv, and aggressive behavior o What’s the extra variable... and what is it?   Lenient parents, and it’s a Third variable Viewing violent television is associated with aggressive behavior very strongly among teens but less strongly among young adults o Age is your Moderator Mediator, third variable, or moderator? the answer: third variable answer is: moderator mediator Multivariate designs and four validities (beta)  Construct o How well were these variables measured?   How reliable, how valid External o To whom or what can we generalize?   To other people? Think about how sample was selected Internal o What other 3rd variables you could control for?   Statistical o How strong and significant are the effects?  Magnitude and statistical significance Chapter 9 clicker:  Which of the correlations in a longitudinal design is MOST important for establishing temporal precedence? o Cross lag  Which of the three criteria for making a causal claim can BOTH longitude designs and multiple regressions designs satisfy. o Covariation Chapter 9 book:  In a cross lag correlation, a measure recorded earlier is associated with different variable measured later. In a cross sectional correlation, two measured variables recorded at the same time are associated. If one variable measured earlier is associated with the measurement of the same variable taken later, this is called can autocorrelation. These three types of correlations are usually found in longitudinal designs.  Multiple regression can help eliminate internal validity problems by controlling for potential third variables. It also sets a predictor variable that acts as the independent variable, or the variable that will influence another variable; and a criterion variable that acts as a dependent variable or the variable that is hypothesized to be affected by another variable. Effect sizes in multiple regression analyses are represented by beta, which is a standard effect size, and b, which is an unstandardized effect size  Multivariate correlational designs are often used when it is impossible to conduct an experiment Clicker:  Parsimony is useful because of… o Its ability to point to a single cause REVIEW FOR EXAM:  In order to account for _____, Dr. Dreyer includes reversed worded items in their questionnaire. This will increase the ____ validity of their results. o Yea saying, construct  Jane purchased a variety pack of a new brand sports drink. They had 4 flavors, she took a sip of berry bonanza and did not like the taste. She concluded that she did not like that brand of sports drink. Why might janes conclusion about that particular brand of sports drink be incorrect? o Janes sample was unrepresentative because she only tried one flavor.  When graphing bivariate correlations, we use a ___ for ___ data, which means very person is represented as a dot. o Scatterplot, quantitative (the “why” )  Which of the following statements regarding mediators is true? o A mediator variable is internal to the causal variable  He finds a strong positive relationship between coffee consumption. What is a likely statistical result based on the description of his results? o R=.56, p=.001  Professor Langston wants to know which NFL teams are supported by Tucson residents. He asks three members of his lab to go out and collect data. Ava goes to her apartment and asks her 3 roommates which NFL team they support. She then asks them to contact friends and family members in the Tucson area who might be willing to get in touch with her about their favorite NFL teams. Ava is using:_____. Brandon walks to the student union to get lunch and decides to ask everyone he passes which NFL team they support. Brandon is using: ______. Kira goes to a local wing restaurant and asks people who are watching any of the televised football games which NFL team they support. Kira is using: ____ o Snowball sampling; convenience sampling; purposive sampling.  Rafael is looking at the multiple regression results from a study predicting sprinting speed from age and stride length. In this example, what is/are the predictor variables o Age and stride length   Predictor= IV (age and stride length)  Criterion= DV ( predict sprinting speed) When trying to determine the relationship between two variables from a bar graph in a bivariate correlation, we compare… o The group means of the categories  What is this an example of: “please select the statement that best describes you”… i prefer watching tv over listening to music, or I prefer listening to music over watching TV. o Forced choice question  When correlating wellbeing and number of friends she finds a strong positive correlation. When correlating number of friends and hours spent with a pet, Dr. Hunting finds a weak negative relationship. When correlating wellbeing and hours spent with a pet, she finds a zero relationship. Which relationship is she most likely to check for curvilinearity? o Number of friends and hours spent with a pet  Which of the following statements regarding beta is false? o The strengths of betas from one regression table can be compared to the strengths of betas from another regression table  Which correlation would Nessa be most interested in if she wanted to know whether increased physical activity led to higher academic performance? o The cross lag correlation between physical activity at time 1, and academic performance at time 2.  Physical activity must come first, physical activity indicates academic performance o What is true for p and r  P tells us about the probability of the result occurring by chance alone or if there is an actual association in the population our data came from, while r tells us about the strength of a relationship. o Which of these samples is more likely to be a representative of a population of 60,000?  A simple random sample of 250 o Adhd one half got red dress girl, other half of class got told blue dress girl When rating behavior of both. This is an example of  Expectancy effects o Longitudinal designs and multiple regression designs each meet two criteria for making a causal claim. Longitudinal designs meet ____ and ____. Whereas multiple regression designs meet ____ and ____.  Temporal precedence and internal validity; covariance and temporal precedence.