[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: 3 Level Nested Anova with equal sample sizes?
Hi Stat et al,
We are trying to establish what analysis to use for the following project. We are testing the reliability between three different 3D digital full-body scanners(machines), their respective anthropometric measurement software (each comes with its own), and
direct anthropometric measurements. We intend to measure and compare digital images from each machine on each type of software and compare them all to the direct measures.
Therefore, we have the following:
Two kinds of measurers (2): machine and person
Three kinds of machines (3): 3 machines
Three kinds of software (3): 3 different programs
N = 40-50 subjects, based on power analysis.
Approximately 30+ anthropometric measurements.
Regarding the idea of a three-level nested ANOVA. What I don't understand how to account for is the comparison the person measurements (direct anthropometry). Removing the person, a simple 2-level nested ANOVA falls out and it's straightforward to
However, we'd like to be able to examine the variance of every software on every kind of scanner along with the direct anthropometry. In turn, we'd also like to be able to examine individual measurements from the scanners and software and compare them to
individual direct measures.
I had thought we could do the following 2 analyses: (1) Run a 2-level nested ANOVA on scanners and software and then (2) Run three coefficients of partial correlation (we have to account/control) for body proportion, etc) on groups of software
measurements by scanner and the direct anthropometry.
We have the follow table drawn up with this in mind for the first analysis (1):
Machine (scanner A,B,C): A B C
Software (a,b,c): aA bA cA aB bB cB aC bC cC
For the second analysis (2), we intend to calculate three coefficients of partial correlation using each group of measurements (aA,bA,cA), (aB,bB,cB), (aC,bC,cC), our calculation of body proportion, and the direct anthropometry.
We're not sure if this is the best method for analyzing the data, and wanted some input and hopefully some kind of blessing. J
Thank you all in advance for any time, effort, and consideration!
Eric Lauer and Co.
* For searches and help try: