any way but Bayes?
0 views
Skip to first unread message
Troels Ring
Dec 7, 2009, 1:56:35 PM12/7/09
to medstat
Dear friends, I have results on 8 pig half-brains which are a bit
troubling. First column is wet weight which is reliable to 1 decimal.
Next column is water content determined with deuterium, also believed to
be reliable to at most a few percent. The third column is then the water
as percentage of the whole weight. This is also decent, but then the
third column expresses the water as a percentage of whole weight minus
water as a measure of dry matter. This is what we are really after but
it is not decent as appears. I can easily convince myself that the dry
matter per 100 g wet weight is common for these animals having been
exposed to the same treatment and then make a Bayesian analysis forcing
this to be true and if I require the variance to be small I can shrink
the last column as much as I please. But who will that convince? I think
I can find some data to make a decent prior but not with water measured
by deuterium so it is still shaky ground. What would you suggest?
Best wishes
Troels
Data
> DEUTERIUM
weight VCNS VCNSww VCNSdw
1 34.65 30.20185 87.16262 678.9753
2 42.28 34.70947 82.09429 458.4812
3 36.49 33.08518 90.66917 971.7159
4 28.19 26.39688 93.63916 1472.1202
5 31.35 28.06814 89.53156 855.2520
6 40.42 35.27202 87.26377 685.1616
7 39.09 32.90085 84.16693 531.5893
8 31.92 27.18567 85.16813 574.2237
troubling. First column is wet weight which is reliable to 1 decimal.
Next column is water content determined with deuterium, also believed to
be reliable to at most a few percent. The third column is then the water
as percentage of the whole weight. This is also decent, but then the
third column expresses the water as a percentage of whole weight minus
water as a measure of dry matter. This is what we are really after but
it is not decent as appears. I can easily convince myself that the dry
matter per 100 g wet weight is common for these animals having been
exposed to the same treatment and then make a Bayesian analysis forcing
this to be true and if I require the variance to be small I can shrink
the last column as much as I please. But who will that convince? I think
I can find some data to make a decent prior but not with water measured
by deuterium so it is still shaky ground. What would you suggest?
Best wishes
Troels
Data
> DEUTERIUM
weight VCNS VCNSww VCNSdw
1 34.65 30.20185 87.16262 678.9753
2 42.28 34.70947 82.09429 458.4812
3 36.49 33.08518 90.66917 971.7159
4 28.19 26.39688 93.63916 1472.1202
5 31.35 28.06814 89.53156 855.2520
6 40.42 35.27202 87.26377 685.1616
7 39.09 32.90085 84.16693 531.5893
8 31.92 27.18567 85.16813 574.2237
--
Troels Ring
Heliosvej 12
9210 Aalborg SØ
98140582
30480176
job: +4599326629
Reply all
Reply to author
Forward
0 new messages