Please enable JavaScript.
Coggle requires JavaScript to display documents.
SEM Multicollinearity Paper - Coggle Diagram
SEM Multicollinearity Paper
Where is the threshold for variables being too correlated when using SEM?
[lm is |r| > 0.7, Dormann et al 2012]
What is the main message of the paper?
In SEM, Multicollinearity being an issue or not, is context dependent (ie it's not as simple as if variables are correlated with an |r| = 0.7 or higher, it's an issue!
It is dependent on...
Sample Size
small n hides the problem
Magnitude of Path Coefficients
Variance of exogeneous
and endogeneous variables
Number of mediators???
Number of predictors???
the more var's the more than likely you will have collinearity (but context dependent)
Latent Variables?
Working Abstract
Why are you doing this?
To better understand and quantify when collinearity is problematic for estimating parameters in SEM
To discover/highlight what potential issues may arise and to what extent they impact on parameter estimates
To provide guidance to researchers using SEM when collinearity is an issue, and how to best rectify/work with it
What did you do?
We tested the accuracy of SEM parameter estimates under different data/modelling scenarios
What did you find?
We found that
What does this mean?
What is it good for?
What is the Paper's Title?
Option 1: Collinearity in Structural Equation Modelling (SEM): If and When is it an Issue?
Option 2
Option 3
What are the figures that I'll use to demonstrate my results in the most succinct and engaging way?
Decision Tree (Choose your own adventure)
Paper 1
RShiny App
Paper 2 or 3?
Diagnostic and Correction?
Can we reverse engineer the problem? Can we determine/quantify the attenuation/inflation of parameter estimate?
Paper 2
Predictability?