RANDOM RAMBLINGS

1. GLASS DATA

I wanted to use the glass data from Quebec, I really did. Unfortunately,
none of the information in the spreadsheet is useful to my cause. There
is no data relating to the number of glass fragment groupings found at
a crime scene, nor is there any data relating to the number of glass
fragments per glass grouping found at crime scenes.

2. EVETT 1994 AND DATA MANIPULATION

We spoke about "maniplulating data" during our last meeting. This was
much more ambitious than I originally expected. Any fool can add random
numbers onto data, but I decided that I should seek guidance to manipulate
it in a meaningful manner.

I decided it would be best to modify the Poisson distribution that calculates
the T values used in Evett's 1994 paper. I chose this because Evett states:
	"Imagine that, before starting the case, the expert was asked how many 
	matching fragments would be expected to be found if C were truly the case 
	and that the reply was 'about 4'."

Clearly there is a lot of wiggle room in the calculation of this factor, so I
created a procedure which writes to a file which can be plotted in gnuplot.
The plot calculates the T values based on finding x fragments in y different
groups with a lambda factor of y. This procedure outputs the T values given
the input values (x,y,z) and can be optained by executing (create-plot-t) after
loading evett94.lisp

The resulting plot was not significant to me in any way. The slopes were poor
for skiing: low values of (x,y,z) produced a local maximum, but as (x,y,z) 
became larger, the outputs became flat.

Since I had created a function to calculate t-values given a lambda and a j
value, it seemed natural to plot lambda, j, and the result of Evett's 1994
algorithm with the inputs lambda and j.

The resulting plot is fairly interesting, but nothing abnormal is present.

3. EVETT 1994 SUMMARY

This summary is available as summary.text