1 comparing sem to the univariate model data from grace and keeley (2006) ecological applications
TRANSCRIPT
![Page 1: 1 Comparing SEM to the Univariate Model data from Grace and Keeley (2006) Ecological Applications](https://reader036.vdocuments.us/reader036/viewer/2022081602/551a8ed85503466b3a8b4e1d/html5/thumbnails/1.jpg)
1
Comparing SEM to the Univariate Model
data from Grace and Keeley (2006) Ecological Applications
![Page 2: 1 Comparing SEM to the Univariate Model data from Grace and Keeley (2006) Ecological Applications](https://reader036.vdocuments.us/reader036/viewer/2022081602/551a8ed85503466b3a8b4e1d/html5/thumbnails/2.jpg)
2
A Graphical View of the Univariate Model
![Page 3: 1 Comparing SEM to the Univariate Model data from Grace and Keeley (2006) Ecological Applications](https://reader036.vdocuments.us/reader036/viewer/2022081602/551a8ed85503466b3a8b4e1d/html5/thumbnails/3.jpg)
3
Initial Univariate Results
ns
ns
We might us a variety of criteria to decidewhich paths to retain. Here we use t-tests.
![Page 4: 1 Comparing SEM to the Univariate Model data from Grace and Keeley (2006) Ecological Applications](https://reader036.vdocuments.us/reader036/viewer/2022081602/551a8ed85503466b3a8b4e1d/html5/thumbnails/4.jpg)
4
Pruned Univariate Model
![Page 5: 1 Comparing SEM to the Univariate Model data from Grace and Keeley (2006) Ecological Applications](https://reader036.vdocuments.us/reader036/viewer/2022081602/551a8ed85503466b3a8b4e1d/html5/thumbnails/5.jpg)
5
What are the causal relationships?Structural Equation Meta-Model
(SEMM)
SpeciesRichness
StandAge
FireSeverity
PlantAbundance
LocalAbiotic
Conditions
Within-plotHetero-geneity
LandscapePosition
Local ConditionsLandscape Conditions
Good time for thought experiments!
![Page 6: 1 Comparing SEM to the Univariate Model data from Grace and Keeley (2006) Ecological Applications](https://reader036.vdocuments.us/reader036/viewer/2022081602/551a8ed85503466b3a8b4e1d/html5/thumbnails/6.jpg)
6
Our Structural Equation Model
![Page 7: 1 Comparing SEM to the Univariate Model data from Grace and Keeley (2006) Ecological Applications](https://reader036.vdocuments.us/reader036/viewer/2022081602/551a8ed85503466b3a8b4e1d/html5/thumbnails/7.jpg)
7
SEM Results
Are these results easier to interpret thanthose from the multiple regression?
![Page 8: 1 Comparing SEM to the Univariate Model data from Grace and Keeley (2006) Ecological Applications](https://reader036.vdocuments.us/reader036/viewer/2022081602/551a8ed85503466b3a8b4e1d/html5/thumbnails/8.jpg)
8
Some of the Amos Output
here we see indications,in the form of p-values,that all parameterscontribute significantly to the model.
![Page 9: 1 Comparing SEM to the Univariate Model data from Grace and Keeley (2006) Ecological Applications](https://reader036.vdocuments.us/reader036/viewer/2022081602/551a8ed85503466b3a8b4e1d/html5/thumbnails/9.jpg)
9
But wait, is the model sufficient?
ask for residualsandmodificationindices, then rerun the model
Model chi-square (p = 0.057) suggests that model is marginally adequate. But, we should perform some sensitivity tests by looking for indications of poor fit and evaluating some alternatives (to be safe).
![Page 10: 1 Comparing SEM to the Univariate Model data from Grace and Keeley (2006) Ecological Applications](https://reader036.vdocuments.us/reader036/viewer/2022081602/551a8ed85503466b3a8b4e1d/html5/thumbnails/10.jpg)
10
What do modification indices say?
MI values greater than4 are suggestive, but thesevalues are only very approximate "hints" ofwhether modifications tomodel would lead to acceptance of additionalpathways.
All these MIs indicate that there may be a significantresidual correlation between heterogeneity and total cover. We might want to see if there is a significantresidual correlation between the two and, if so, to consider what process that would represent.
![Page 11: 1 Comparing SEM to the Univariate Model data from Grace and Keeley (2006) Ecological Applications](https://reader036.vdocuments.us/reader036/viewer/2022081602/551a8ed85503466b3a8b4e1d/html5/thumbnails/11.jpg)
11
What do residuals say?
residuals ambiguous?.
![Page 12: 1 Comparing SEM to the Univariate Model data from Grace and Keeley (2006) Ecological Applications](https://reader036.vdocuments.us/reader036/viewer/2022081602/551a8ed85503466b3a8b4e1d/html5/thumbnails/12.jpg)
12
Try alternative model
chi-square drops from 20.60 to 13.39, that's a difference of 7.21, indicating a significant improvement to the model.
![Page 13: 1 Comparing SEM to the Univariate Model data from Grace and Keeley (2006) Ecological Applications](https://reader036.vdocuments.us/reader036/viewer/2022081602/551a8ed85503466b3a8b4e1d/html5/thumbnails/13.jpg)
13
Now we are ready to consider the results!
our unstandardized estimates
our standardized estimates
![Page 14: 1 Comparing SEM to the Univariate Model data from Grace and Keeley (2006) Ecological Applications](https://reader036.vdocuments.us/reader036/viewer/2022081602/551a8ed85503466b3a8b4e1d/html5/thumbnails/14.jpg)
14
More results
covariance betweenheterogeneity andcover is significant.
![Page 15: 1 Comparing SEM to the Univariate Model data from Grace and Keeley (2006) Ecological Applications](https://reader036.vdocuments.us/reader036/viewer/2022081602/551a8ed85503466b3a8b4e1d/html5/thumbnails/15.jpg)
15
And Still More Results
R2 for richness is pretty good, another indicator of model adequacy.