Partial residual graphs for linear regression
---------------------------------------------
by Joseph Hilbe, Editor, STB, FAX: 602-860-1446


            ^partres^ depvar varlist [^, lowess^]

produces partial residual graphs for each independent variable in varlist.
The program implements the ^lowess^ option by invoking ^ksm^, so you must
install gr6 for this option to work.  ^lowess^ graphs a lowess smoother
over the partial residual graph, allowing the user to determine more
easily the shape of partial residuals.  With or without the lowess curve,
the graphs aid in detecting nonlinearities as well as identifying cases
with high-residual values.

Each variable is calculated separately and is not standardized. The formula
for partial residuals is:

            r = Xij*Bij + (Yi - Mi)

where (Yi- Mi) are the values of the residuals produced as a result of
linear regression of Y on Xi. The partial residual graph is a 
scatterplot of the partial residuals on the y-axis by the actual 
variable values on the x-axis. Unlike the ^leverage^ command, 
^partres^ displays all independent variables automatically.

Be certain that ^ksm.ado^ is in the same directory or path as ^partres^;
otherwise the ^lowess^ option cannot be run.