Nested design help in JMP 9 for Repeated Measures Data

[#1]

I'm trying to analyze the following repeated measures data:

I have 6 subjects (ID 1-6)

Each subject is tested at 4 levels (45,55,65,75)

and 6 frequencies (1-6)

The dependent measure is "weight"

So my JMP Table has the following Columns:

Subject ID, Frequency, 45 Weight, 55 Weight, 65 Weight, 75 Weight

There are 36 columns (6 sub x 6 frequencies and 4 weights across)

I am using the MANOVA platform with the 4 levels as the responses and frequency as the model effect.

I am then using the "repeated measures" in the response specification naming the Y the "level"

Is this the correct way to do this analysis?

The research question I am trying to answer is if there is the weight changes depending on level, frequency and a level * frequency interaction.

I need to use repeated measures because I have the same subjects making multiple response across sampling (which in this case is level).


I've also tried the regular fit model with nesting random effects but I can't seem to get that to work correctly. (I've stacked the level column in this case)

I am doing:
Subject[Level, Frequency]&Random
Frequency
Level
Frequency * Level

Any help would be much appreciated.

Dan

[#2]

In this case,

I will write model at Construct Model Effects of "Fit Model".
Because Frequency is nested from each combination of Level and Subject.

Level
Subject&Random
Level*Subject&Random
Frequency[Level,Subject]&Random

[#3]

Dan,

How was the experiment conducted. That may help determine the right model.

e.g. Did you randomize the order of any of the factors (frequency or level). Or was every subject given the same order of factor combinations?

Also, can you give some context for the variables? What does the response "weight" represent. What is the research question you are trying to answer?

[#4]

We had each subject listen to 1000 trials per level composed of 6-tone complexes using a 2 AFC loudness paradigm. Subjects voted for the louder interval. For these conditions we were using levels of each 6-tone (6 frequency) complex randomly chosen from a gaussian distribution of levels +- 15 dB from the mean level.

We ran ascending levels for half the subjects and descending for the other half assigned randomly.

Using the 1000 trials we determined the relative "beta" weights by correlating the subject's response (which interval was judged as louder) with the relative local level difference between the levels of each tone. This analysis was a multiple regression, that is based on previously published experiments in our lab.

We determined the weights for each tone in the tone complex per level for each subject and level.

The research question is does the pattern of loudness weights change as a function of frequency, level, or the interaction of frequency by level.

frequency is nested in level.
Each level had the same 6 tones.
Each subject was given the same combination of conditions.