[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]
Re: [GSP1954@aol.com: MORE DINOSAUR ENERGETICS] (fwd)
From: Johnathan Melville <firstname.lastname@example.org>
> To anyone interested in my two cents please look this over. ...
> However I do not see a problem with the methods or sample
> size that Ruben Geist and Jones used in thier paper.
I tend to agree with Dr. Paul here. A sample size of four is almost
always inadequate, for almost any purpose. The only time it is
acceptible, IMHO, is when there simply aren't any more samples
available. And there are certainly more species of birds available
BTW, my last degree is a Master of Science in Biostatiatics, so, on
this subject, at least, I can claim a certain expertise.
> Hello, my name is J.Melville and I am a neuro-physiologist here at
> OSU. I just wanted to state for the record that although the data set
> presented by Ruben Jones and Geist may be some what small and may
> possibly under represent the diversity of endo and ectothermic animals,
> one fact remains. There is a four fould difference in crossectional
> area of the nasal cavities of these animals.
The question is whether this difference would remain if a larger sample
were used. In short, it could easily be a sampling artifact due to
the small smaple size.
Also note that in evaluating the importance to RT's as a marker of
basal metabolism the *variance* of the RT size is as important as the
average diffeence. If the ranges of the sizes in the two groups overlap
substantially, then RT sizes are not necessarily indicative of thermo-
regulatory strategy. And the variance simply CANNOT be determined
accurately from a sample size of four.
> This is based on very sound statistical methods of multiple linear
Yep, a known statistical method, one that usually requires a *far*
larger sample size than eight or ten for accurate results. [For
instance, for the related method of ANOVA, the results are considered
suspect if any category has less than five measurements in it; and
MLR has even more stringent requirements than ANOVA, as it is trying
to extract more information than ANOVA does].
> The model that was calculated for the two lines that they recieved
> described over 95% of the variance in thier data set and the
> difference between the two lines was highly significant.
95% of the variance if a mere handful of specimens!?!?!? That is
hardly any variance at all. The question is how much lower would this
number be if an adequate sample size were used.
> This four fould difference does exist in this data set!!!
Yep, so ratites have a fourfold larger RT's than lizards.
Big deal :-}
Now, do *birds* have fourfold larger RT's than lepidomorphs! The sample
used doesn't even *address* this issue. To get at that question one
needs a broad sample of birds from a wide variety of clades, plus
a substantial variety of lizards, snakes, and tuataras.
> So it is quite posiible that through other anecdotal evidence that
> others may disagree with thier conclusions but THIS DIFFERENCE IS
For *that* *sample*. And since the avian part of the sample is mostly
ratites, all it really says is that *ratites* proabably have large RT's.
> All of these criticisms should have worked against thier data set!!! Not for
> it. Lest this is unclear to anyone I will explain this briefly. Each point on
> a regression ususallly represents the average of a population of
> responses. ...
> However, with only one response at any given point in
> time, you most likely would not get the best linear fit.Thus, you would want
> several trials or several different people to run for an hour and take the
> same measurements and then plot this data on top of the data you originally
On the other hand, the smaller the sample, the more likely it is to
get an accidental pattern, one that is only real for the actual sample,
not for the population as a whole.
[Compute the standard errors of their numbers].
The peace of God be with you.