Some basic questions about my first DOE response surface design

[#1]

Hi everyone,

I'm quite new to JMP Pro and the questions I'm about to ask are pretty basic. 

I have two independent variables and a dependent variable (response), all three measured using a seven-point Likert scale. The goal is to analyze the within-individual perceptual (in)congruence between the independent variables and their influence on the response. My sample size is 260.

Question 1: When I click on "make table" before adding data to the columns, I get the following output. I have some idea what the column Pattern means but I don't know if I should keep it in the analysis. When I populate the 260 records into the X1, X2, and Y columns, what should I do the with Pattern column? Should I keep it as is (with only 10 values while X1, X2, and Y have 260 values) or just get rid of the column?

Question 2: Did I correctly determine the values for X1 and X2 here? Should I do anything about the Lower and Upper Limit for the response as well (i.e., Lower Limit = 1; Upper Limit = 7). 

Question 3: In my case, what should be the number of center points? The default I'm seeing is 2. 

Your help is greatly appreciated. Thanks.

[#2]

First, welcome to the community.  Unfortunately, you have not provided enough context about your situation to provide specific advice.  I do have the following questions/comments:

• Sample size of what?  Is this a product or a web page or what?  What is being sampled?
• What are the independent variables?  What is the response variable?
• You apparently already did the "study"?  Is there more than one person in the study? If so, you should have a column identifying the individual(s).
• I'm confused the statement these are on a 1-7 scale.  I assume you mean Ordinal scale.  Likert is a specific type of ordinal scale (invented by Rensis Likert).  It uses words like disagree, somewhat disagree, etc. about statements.  Can you get more than one number for any measure of the sample?
• It seems to me you question the measurement system?  Likely both within individual and between individual components of variation.  This is good to question.
• It does not appear as though you are running an experiment.  So I think you are in the wrong platform?
• Perhaps you are more interested in MSE and possibly Kappa (for inter reliability of categorical data)?

https://www.jmp.com/support/help/en/17.2/?os=mac&source=application#page/jmp/overview-of-attribute-g...

https://www.jmp.com/support/help/en/17.2/?os=mac&source=application#page/jmp/agreement-statistic-rep...

Oh and you can likely ignore the pattern column.

[#3]

Dear @statman 

Sorry about the confusion.

Sample size of what?  Is this a product or a web page or what?  What is being sampled?

The dataset is a survey of 260 employees about their perceptions of three variables. So, n = 260. As I mentioned in the original post, these are within-individual perceptual measures. This is behavioral research using perceptual measures. 

What are the independent variables?  What is the response variable?

The independent variables are employees' perceptions of two competing factors regarding the organizational culture. The dependent variable is the employees' level of compliance with corporate policies. All three variables were measured the same seven-point, agree/disagree, Likert scale. 

You apparently already did the "study"?  Is there more than one person in the study? If so, you should have a column identifying the individual(s).

Yes. Originally, I analyzed the data using structural equation modeling and submitted the paper to a journal. The editor requested using polynomial surface analysis so that I can examine the actual (in)congruence between the two competing perceptions and they're joint relationship with the response (i.e., compliance level).

I'm confused the statement these are on a 1-7 scale.  I assume you mean Ordinal scale.  Likert is a specific type of ordinal scale (invented by Rensis Likert).  It uses words like disagree, somewhat disagree, etc. about statements.  Can you get more than one number for any measure of the sample?

Yes. I'm expected to treat the Likert scale measures as continuous variables. During the survey, a respondent was able to choose a number from 1 (strongly disagree) to 7 (strongly agree) for each of the three variables. For example, one employee could choose 3 for one independent variable, 4 for the other independent variable, and 5 for the response or dependent variable or could choose the same number for all three. 

Hope it makes more sense now.

[#4]

Just a quick follow up. You have 3 response variables. Your input variables are within and between individuals. See my previous advice MSE and Kappa

[#5]

@statman 

No, I have one response variable (Y) and two independent variables (X1 and X2). There is no between-individual data. Each of the 260 respondents provided a rating (from 1 to 7) for the three variables. 

P.S. I looked into MSE and Kappa. I'm required to use the response surface methodology and produce the 3D surface graphs.

[#6]

I'm so sorry.  I don't understand the situation or lack the appropriate SME to help you.  Hopefully another member can provide assistance.