Re: Simple Statistics Question

On 13 Oct 2005 00:43:21 -0700, "Colin E." <colin.eberhardt@xxxxxxxxx>
wrote:

> Dear All,
>
> It has been a long time since I studied statistics. I was wondering if
> someone could point me in the right direction regarding which
> statistical technique I should be using for the following problem ...
>
> At work we are currently trying to optimise our business practices. We
> have measured the time we spend in a number of activities, then changed
> practices and re-measured.
>
> I have data which resembles the following:
>
> For 3809 hours of effort spent on a project we spent 5.7% of our time
> on Activity A.
>
> After changing our practices, out of 13395 hours of effort we spent
> 2.5% of our time on Activity A.
>
> What test should I use to determine whether the our change of practice
> has had a significant effect (within a certain confidence level).

What are the units that are independent and meaningful?
What units of counting can be the basis for a 2x2 contingency
table?

"3809 hours" and "13395 hours" are large numbers to start
with. My difficulty here is that I find it hard to accept that
a task or an assignment corresponds to one hour, always
and ever. In fact, it seems likely that "one hour" may be
a minimum "activity", as written on time sheets.

If tasks are assigned by *week*, then this corresponds to
5 out of 95 for the early experience, compared to
8 out of 335 for the latter -- This is a contrast that would
not be "statistically significant" at the 5% level, giving
a chi-squared of 2.08. However, if there are independent
assignments by *day*, the test value would be 5 times
as large, readily supporting the idea that a systematic
difference was created....

The data might "support the inference," but this is still
a weak inference unless you can rule out other sources
of change. It would help if you could measure the change
in more detail.

I do not have experience in monitoring jobs, but from what
I have read, I expect that a better design would try to count
the number of incidents (or whatever), and also measure their
duration. With both numbers on hand, you could draw on
common sense when you argue that the change-of-practice
measurably decreased either the count of A, or duration of
A, or both.

>
> If someone could just point me in the right direction I will happily
> run of to the library to refresh my memory on the subject!

Hope this helps.

--
Rich Ulrich, wpilib@xxxxxxxx
http://www.pitt.edu/~wpilib/index.html
.