Slitherine

quote:

ORIGINAL: FaneFlugt.DK

How you use a sheep in your economy is offcourse open to debate.

Im not claiming sheeps are the best animal to study. Unfortunatly its hard to come across a good collection of say 50 whole boars that are tagged with age, sex. But the zoological museum in Copenhagen had a collection of sheep. So i had to use those.

The study is purly methodic, is the method valid?

What I would raelly like is 100 dead boars with know sex and age. Then do a sex, age study and compare the study. To a neolithic slaughter site. To see the sex, age distribution of the site, It would be cool to see if people back then, where selective hunters or overkillers. It would show up as overkill if there where to many female boars in the bone material.

One can offcourse also debate, if a sex difference in sheep. Is applicable to other large Artiodactyla. But Im just a student doing a scientific method paper

quote:

ORIGINAL: JAMiAM

You may need to establish several conditions, before you get started. What do we know about the existing samples of sheep bones? Do we know that all of them came from animals that were the same age, before death? If bmd is a function of age, then this may be an important factor. We also need to know (or assume) the distribution of the sex (male/female) in the population of bone donors, and if age was a concern, then the distribution of their ages.

I'd say that you then need to determine that there is a significant statistical difference between the mean bmd(x) and mean bmd(y). This can be done by using a difference of means test. Assuming that the data supports the hypothesis that there is, according to the confidence interval that you've chosen, then you'll need to figure out some other things.

What is the distribution of bmd within a population? That is, what distribution best describes the random variables bmd(x) and bmd(y)? Generally, in cases like this, a Normal Distribution is assumed, with the mean and standard deviation being taken from the sample. If this can then be assumed, and that the mean bmd(x) does not equal mean bmd(y), and that P(f) = P(m), and that donor age is not a significant factor (whether by insignificant variance in age of donors, or by insignificant variance in bmd, due to age), then what I would do is to figure out what the confidence intervals for the combined superimposed set of Normal Distributions would be, with a line dividing the two non-intersecting C.I.'s at [(mean bmd(x) + mean bmd(y))/2]. Note, that given equal standard deviations between x and y, you should have equal area in each of the C.I.'s but they will not be symmetrical about their respective means. The calculation of the area under the C.I's will give you the probability that a bone falling within that range will be either a male, or a female.

It may very well be that you can refine this further, dependent upon the difference in mean bmd(x) and mean bmd(y), or by establishing some conditional probabilities concerning the overlaps on either side of the centerline. However, the method I gave you, should be replicable by a first year statistics student and "good enough for government work"...

Hope this helps.

quote:

ORIGINAL: bjmorgan
My guess is that preparing confidence intervals would be confounded by large standard deviations relative to the means. That basically means that the differences of BMD within a sex might be large relative to the differences between the sexes. At that point, you'd end up with probabilites of being male or female that are too similar.

quote:

ORIGINAL: JAMiAM

quote:

ORIGINAL: bjmorgan
My guess is that preparing confidence intervals would be confounded by large standard deviations relative to the means. That basically means that the differences of BMD within a sex might be large relative to the differences between the sexes. At that point, you'd end up with probabilites of being male or female that are too similar.

This was my main concern, as well. C.I.'s of 85% or greater would reasonable for his work. If they turned out to only be 60-80%, then much less so.

Another concern is that if the samples he possesses are of recently slaughtered animals (last few decades), then there needs to be a method for estimating the decay of BMD of the fossils, so that a realistic estimate of BMD at time of death could be used. Not to mention other issues like differences in grazing methods leading to variation in BMD, etc.

quote:

ORIGINAL: bjmorgan

quote:

ORIGINAL: JAMiAM

quote:

ORIGINAL: bjmorgan
My guess is that preparing confidence intervals would be confounded by large standard deviations relative to the means. That basically means that the differences of BMD within a sex might be large relative to the differences between the sexes. At that point, you'd end up with probabilites of being male or female that are too similar.

This was my main concern, as well. C.I.'s of 85% or greater would reasonable for his work. If they turned out to only be 60-80%, then much less so.

Another concern is that if the samples he possesses are of recently slaughtered animals (last few decades), then there needs to be a method for estimating the decay of BMD of the fossils, so that a realistic estimate of BMD at time of death could be used. Not to mention other issues like differences in grazing methods leading to variation in BMD, etc.

much money. And now I've realized that statistics is much more fun.)(No, really.)

quote:

..more fun than actually doing the field studies in mountainous regions of far flung isolated areas ..