Search code examples
rr-lavaansemplot

How to interpret path coefficients in path analysis


I am trying to come up with a path analysis diagram using lavaan and semPlot.

1) Does anybody know how to interpret the path coefficients, especially those that does not originate anywhere but points to themselves, for eg. the coefficient 1.00 pointing to "DW", coefficient 0.61 pointing to "ANTS" and coefficient 0.92 pointing to "RH"?

2) Any idea how I can customise the graphics in the diagram, like add a title or make the box bigger? I was referring to this site (http://sachaepskamp.com/documentation/semPlot/semPaths.html), but it does not show any example of adding titles and I've been trying out different codes with no success.

3) What information should I present? Is it sufficient to show the p-values of all the pathways and indicate which is significant? Thank you!

Specify Model for ants diversity

Model.ants<-'Ants~AGB+Max.Temp+Min.RH+Deadwood.quantity
Deadwood.quantity~AGB
Max.Temp~AGB
Min.RH~AGB
Max.Temp~~Min.RH'

Results

Result.ants<-sem(Model.ants,data=combined)
Fit.ants<-cfa(Model.ants,data=combined)
summary (Result.ants,standardized=TRUE,fit.measures=TRUE,rsq=TRUE,
modindices=TRUE)

Plot path

semPaths(Result.ants,"std",intercept=FALSE,edge.label.cex=0.8, 
curvePivot = TRUE,layout = "spring",
nodeLabels = c("ANT","DW","Temp","RH","AGB"),title = TRUE)

Path Analysis Diagram


Solution

    1. Please refere to to semPlot: Unified visualizations of Structural Equation Models:
    • Directed edges indicate linear regression parameters
    • Bidirectional edges indicate (co)variances
    • (Residual) variances can be indicated in (...) Double headed selfloops (RAM style).
    1. You can add title to a chart by calling title function and change size of element by assignment the argument sizeMan in semPaths:

      semPaths(fit, "std", layout = "circle", sizeMan = 2)
      title("My Paths")
      
    2. You should presents all the model paraments (regression coefficients, variances, covariances etc.) moreover "p < 0.05 then the effect is significat" does not work for SEM generally.