Hierarchical analysis, means of least-square values
- From: bio <bio@xxxxxxxxxxxxxxxxxxxxxxx>
- Date: Sun, 27 Jul 2008 14:44:37 GMT
My experimental datasource lets me take replicate measures of individual
samples. The first layer of analysis is a linear model in R; the fits
are almost always very colinear (adj. Rsquared >0.99, 1%/99%confidence
int. +/- 0.01). The slope of the fitted line is the value of interest.
The next level of analysis will be to compare slopes from replicate runs
of the same sample. These vary in the 2 to 5 % range, and I would like
to find an appropriate estimate and confidence interval for the repeated
measures of the slope. How can I combine the linear model results and
obtain a "second-order" estimate and confidence interval?
(Since the linear models fit so well, as a practical matter I can treat
the slopes as raw data and fit a normal distribution. But I could not
answer a critical challenge to that method.)
And finally, I will want to measure samples after different treatments
and be able to reject the null hypothesis - no change in slope estimate
due to treatment. That will be a "third-order" comparison of
second-order estimates, this time in an ANOVA-type comparison; ideally
yielding an estimate and confidence interval for difference in slope
given treatment.
Grateful for any suggestions...
Albert W.
.
- Follow-Ups:
- Re: Hierarchical analysis, means of least-square values
- From: illywhacker
- Re: Hierarchical analysis, means of least-square values
- From: RichUlrich
- Re: Hierarchical analysis, means of least-square values
- From: Robert Dodier
- Re: Hierarchical analysis, means of least-square values
- Prev by Date: Polarised sunglasses see black hole disks
- Next by Date: Multidimensional correlation measure?
- Previous by thread: Polarised sunglasses see black hole disks
- Next by thread: Re: Hierarchical analysis, means of least-square values
- Index(es):
Relevant Pages
|
