OT: AW: Loop Survey - Summary - post scriptum to eben

["Rainer Thelonius Balthasar Straschill" <rs@moinlabs.de>]

Eben, you said (replying to my age distribution question):

> I'm not certain I understand your question about years of 
> looping experience, although you may be asking why there are 
> more people in the middle amount of years and less on the 
> extremes, which is how almost all data looks when you take a 
> survey (there is an average with most people in the middle, 
> and less people are further away from average). 

No, the question is more specific.
First of all, of course you're right regarding the age of the loopers (for
which we also see some approximately Gaussian binning), but not regarding
the years of looping experience, and I'm gonna explain why:

The mathematical model for this works like this. Any looper i has started 
at
a point t1,i in time with looping. Now when the survey happens at t2, he
will enter t2-t1,i as a reply to that question (which also has some
predefined bins, namely <1, 1-3, 3-5, 5-10 and >10 to somehow stick with 
the
typical bins of human resource departments).

You may argue that there might be some answers from people who stopped
looping at a time t3,i<t2. But I believe that those wouldn't have answered
that survey, so it's fairly reasonable to assume that all of the loopers
have been looping more or less all the time (at least everyone is still
looping right now, and most of them in all of or a considerable amount of
their current work).

So what I'm looking at: how was the distribution when the people started
looping in the past. If you model that the same number of people started
looping every year, then after 18 years of looping, you get the 42% value 
we
have for the >10yrs people - but then the other bins look differently (the
1-3 and 3-5 bins are about double the size of the <1 bin and the 5-10 bin 
is
nearly three times as big as the 3-5 bin because these bins have different
widths). More precisely, this model gives me values 5.3, 11, 11, 32 and 42,
respectively.

So how has the number of loopers which joined each year in the past have to
look to lead to our distribution of 5 - 15 - 19 - 20 - 42? I did some 
simple
runs with the Solver in Excel, and it looks like it's a distribution with
the first people starting more than 20 years ago, a very slow increase from
then, then some sharp peaks about 9 and 5 years ago and since then a slight
decline.

However, this set of equations is underdetermined, so: with your knowledge
of statistical methods, would you be able to deduct the generating function
or parameter for a surface which describes the solutions to this set of
equations?

        Rainer