The point (31.7, 176.2) was an outlier that was removed for the second set of calculations for this data. You may wish to link to the calculation sheet below. Here is a graph of the data and a linear regression best fit line.

Next the residuals were graphed as well as some upper and lower bounds. Then counts were made to get the 63%, 69% and 100% of data within the three intervals. The interval corresponding to 63% was plus or minus 2.225 CM computed as the average of the absolute values of the residuals. The interval corresponding to 69% was plus or minus 2.670 CM computed as standard error of the residuals. And the interval corresponding to 100% was plus or minus 4.1 CM computed as the maximum of the absolute values of the residuals.

One conclusion drawn from this data is a linear regression line give or take an interval that makes your prediction reasonable for one purpose or another. So the interval:

[ hlow(x) = 4.2 x + 59.3 - 2.2, hhigh(x) = 4.2 x + 59.3 + 2.2 ] =

[ 4.2 x + 57.1 CM, 4.2 x + 61.5 CM ]

may represent a way of predicting an estimate give or take the average residual.

Copyright © 2002 with all rights reserved by William V. Thayer