Design of Experiment:

[Abhishek Shinde]

Hello,

My design of experiment is as follows:

Considering 2 between-the-subject factors with 2 levels each, I have designed 4 different experiments and have recruited 30 subjects for each experimental setup. I have collected 8 responses (within the subject factors) from each individual.

(4 differnt experiments: hAhB hAlB lAhB lAlB, with 30 subjects each)

I have arranged data in following format

Now how should I do the analysis? I am unable to get any reference which deals with such kind of problem. I would be grateful if any of you can help me in this regards.
What should be the Stata command for analysis?

xtmixed Y A B || C: D: E:|| Subj Name: is not working on stata.

[Clyde Schechter]

In your data layout, most of the observations are not attributed to any particular subject. I assume you intend that, for example, observations 2 through 8 all refer to subject hAhB-01, who is named in observation 1, etc. Also, clearly what you are showing us is not a Stata data set because Sr No and Subj Name are both illegal as variable names due to embedded blanks. I'll start by assuming that you have actually imported this data into Stata and that your variables have been given legal names such as Sr_No and Subj_Name.

// IDENTIFY EVERY OBSERVATION WITH A SUBJECT
sort Sr_No
replace Subj_Name = Subj_Name[_n-1] if missing(Subj_Name)
// AND CREATE A NUMERIC SUBJECT IDENTIFIER
encode Subj_Name, gen(Subject)

// MODEL THE RESULTS
mixed Y i.A i.B i.C i.D i.E || Subject:

[Abhishek Shinde]

Thank you very much for a quick reply.

In your data layout, ...................such as Sr_No and Subj_Name.

Yes sir, you have understood my problem correctly .

// MODEL THE RESULTS
mixed Y i.A i.B i.C i.D i.E || Subject:

Isn't there any need to specify that A and B are between the subjects and C,D,E are within the subjects factors in above model?
Or command mixed Y i.A i.B i.C i.D i.E || Subject: is is differentiating within and between the subjects automatically?

Thank you once agian

[Clyde Schechter]

The data themselves identify the within and between effects: A and B will be constant for any given subject but differ across subjects, whereas C D and E will differ within subjects as well as across subjects. Nothing explicit needs to be said about this in the command.

[Abhishek Shinde]

Thank you very much for clarification.

I have a last doubt, I would be grateful if you could help me.

If Y is an ordinal variable taking 3 values (Disagree, Neutral, Agree - encoded as 1,2,3), then how should I go about it?
"xtmelogit " is not useful since I am having observations other than 0 and 1.

[Clyde Schechter]

You could look at -meologit- to do an ordinal logistic regression. Some people would also be comfortable considering this as a continuous variable anyway and sticking with -mixed-.