Path Analysis Remember What Multiple Regression Tells Us

Path Analysis

Remember What Multiple Regression Tells Us • How each individual IV is related to the DV • The total amount of variance explained in a DV • Multiple regression is VERY DV CENTRIC!!! • We know NOTHING about how the IV’s are related

What is Path Analysis? • Another way to do Multiple Regression • A special case of multiple regression where all the relationships among variables can be estimated simultaneously

What do you mean Simultaneously? • Estimate correlations among your IV’s • Estimate regression relationships in the model • You can make all the variables related to one another and find out what the relationship is exactly! • All the regression analyses we did in SPSS can be done in path analysis

Example • Lets say you are interested in the extent to which intellectual activities mediate the relationship between education and cognition

SPSS • SPSS tells you about mediation • SPSS tells you the relationship between – Intellectual activities and cognition – Education and cognition • SPSS doesn’t tell you anything about the relationship between education and intellectual activities

Path Analysis R 2 = ? Education a Activities R 2 = ? b Cognition c All these parameters are estimated at the same time!

• If we wanted to get the same results in SPSS we would have to do a bunch of different models – Education predicts cognition – Activities predicts cognition – Both Predict cognition • If we had more than three variables you can see how this could get impossible

A model like this would be impossible in SPSS IV Mediator IV DV IV Mediator IV IV

Benefits of Path Analysis • Estimate all the relationships among all variables SIMULTANEOUSLY • Able to test various hypothesis regarding different relationships • Most programs you do path analysis in having Missing Data Estimation procedures!!! – SPSS uses list-wise deletion

Statistical Packages for Path Analysis • Path analysis is the precursor to Structural Equation Modeling • Any SEM program can do path analysis – AMOS – LISREL – EQS

AMOS • AMOS is now bundled with SPSS • It is the EASIEST program to use • It’s ease is based on the fact that you actually DRAW your models right in the program!!!

AMOS Steps 1. 2. 3. 4. 5. 6. Open AMOS Tell AMOS what SPSS data set to use Draw your variables in AMOS Select estimation and output options Draw your model(s) Run your model

(1) Open AMOS

(2) Tell AMOS what SPSS data set to use • Select FILE on the top tool bar – Select DATA FILE from the drop down menu Click on File Name and select your SPSS data set

Drawing Tool Bar (3) Draw your variables in AMOS

Draw Your Variables are represented by a box You click on this icon to draw a box in AMOS!

You write the NAME of your variable here You RIGHT CLICK on your variable box to get the “Object Properties”

(4) Select estimation and output options • Select VIEW on the top tool bar – Select ANALYSIS PROPERTIES from the drop down menu

(5) Draw your model(s) age Correlations memory Regression Path

DV needs a way indicate variance not explained by AGE 0, Error_1 1 age memory

AMOS practice • Education and activity engagement are predictors of cognition. Also allow education and activity to correlate – EDUCATION (educ), activity (activ), cognition (cog)