RE: Kurtosis and Skewness

["Jonathan Siegel" <jmsiegel@yahoo.com>]

Vince,

I think it might be helpful to begin by asking how you plan to use your
findings. What do you intend to do with your results? What are the
possible courses of action you can, take and how do you intend for your
findings to affect the way you manage work orders? 

There are several possible approaches depending on what you intend to
do.

It might be helpful to plot a histogram (bar-chart) of the data to give
a visual picture of what percent of the work-orders are hugging the mean
and what percent represent extreme outliers. If you have data for
multiple years, it might also be worth plotting the values and
attempting to discern a trend over time (perhaps do a control chart if
you have enough history) rather than attempting to compare this year to
last in isolation. 

It's worth noting that you'll always have a right-skewed distribution in
time-to-completion situations, simply because zero is the minimum value
(nothing can ever be done in negative time), but there need not be a
maximum value. Simply because times can only be positive, the variation
won't be symmetric, skewness will be positive, and you should expect the
mean to be greater than the median. 

It's also possible that the "extreme outliers" represent a class of
work-orders so different that you'd be best off not managing them
together with the more usual work-orders; you may have two different
management classes. This would be particularly true if you find yourself
in a "tail wagging the dog" situation, e.g. adding or subtracting one or
two outliers creates a large change in the overall mean. 


Sincerely,

Jonathan Siegel
jmsiegel@yahoo.com