regpt.hlp.htm

.-
help for regpt
.-

Regression with a Moving Point
------------------------------

. regpt, [ r(correlation) n(sample size) ]

When you start regpt, you can supply the correlation
between X and Y and the sample size, or accept the default
correlation of .3 and sample size of 30.  Based on the 
correlation and sample size, points are randomly generated
and then plotted showing a scatterplot with regression line.
You are then shown a dialogue box that allows you to move
a "Moving Point", shown as a yellow square (assuming
you are using the default Stata color scheme).

You can move the point in the Y axis, adjusting its distance 
from the regression line without the point included, 
i.e. (Y-Yhat) .  As you adjust the point, you can compare 
the regression line without the point (in green) with the 
regression line with the point (in red).

You can move the point in the X axis.  As you move the point
along the X axis, the distance from the regression line
without the point (Y-Yhat) is held constant.  This allows
you to see the consequences of varying X while keeping 
Y-Yhat constant.

As you move the "Moving Point", you can see the equation for
the regression line with the point change (and compare it to
the regression line without the point).  You can also see 
some Regression Diagnostic Statistics for the regression
equation with the point in the equation, namely the
residual, leverage and Cook's D.

By default, you are shown a plot that has a scatterplot
and regression line.  You can choose to view other plots
  - Scatterplot with Regression Line
  - Residual vs. Fitted Plot
  - Leverage vs. Residual Squared
  - Leverage vs. Residual Squared with Cook's D
    (Cook's D is indicated by the size of the circle).
When you change the Plot type, you can click "Reshow Now"
to immediately switch plot types.

When you are done, you can click the Done button.
You can reset X to 0, and Y-Yhat to 0 via the Reset button.
You can see this help window by clicking Help.


Author
------

     Statistical Consulting Group
     Institute for Digital Research and Education, UCLA
     idrestat@ucla.edu