Instructions for Lab

Using AMOS, fit the Path model on page 125 of Kline after reading in
the data from exerciseillness.csv.  Fit the "just-identified" model,
that is the model with all the arrows included, the degrees of freedom
should be zero.  Note, Kline re-scales the standard deviations
(apparently because there was some computational problem for his software) but it is not
necessary to do this, the unstandardized estimates you get should look
like those I have in the notes rather than the ones Kline has in the
book, the standardized estimates will be identical regardless of
scaling.
============================================

1. Investigate the statistical significance of the direct and
indirect effects.

a.  To get the significance of direct effects, you can look at the
C.R. in the output

b.  To get the significance of indirect effects, you can use the
bootstrap technique.  This technique provides a method for estimating
standard errors for estimators which are complex.  While it is
possible to derive a formula for the standard errors of the indirect
effects (Using the delta method), it is easier to just use the
bootstrap.  Use the parametric bootstrap to find which indirect paths
are significant??

============================================

2.  Drop all direct effect which were not significant and rerun the model.

a. What is the Chi-square testing and how was the degrees of freedom
obtained.  
b. Are there any significant Modification indices for the
model?

============================================

3.  R-squared.  

In AMOS, to get the amount of variability in any outcome variable that
is explained by the variables directly effecting it, look at the
Squared Multiple correlations.  These are output directly to the
screen if you ask for standardized results.  Note the R-squared is
equal to 1-sigma_e where sigma_e is the standardized variance of the
error term associated with the outcome or it can be calculated from
unstandardized results by taking 1 - (estimated variance of error/
implied variance of the outcome variable).

a.  What is the R^2 for Fitness, Stress and Illness for the model you
fit in Q2.

b.  Does the R^2 for Illness change much if you add paths that do not
directly go to Illness?

========================================