Correlation analysis

[robertrao]

I posted this question earlier but didnt get what i needed so reposting it.

It is seen that the rating for Established patients was very high for the questions. ( All the data is not shown)

We waned to see for which "questions" between the two groups ( established or NEW) there was a statistically significant difference???

Please let me know how can we solve this

Thank you

ID              DOMAIN                                                                      QUESTION                                                             SCORE                        APPOINTMENT_TYPE

101             Med Practice: Care Provider                                 Med Practice: CP Concern                                   100                                    Estblished

101             Med Practice: Care Provider                                 Med Practice: CP Explanation                            100                                    Estblished

101             Med Practice: Care Provider                                 Med Practice: CP Discuss treatments              100                                     Estblished

101             Med Practice: Care Provider                                 Med Practice: CP Efforts                                      100                                    Estblished

101             Med Practice: Access                                            Med Practice: Ease of contacting                         100                                   Estblished

101             Med Practice: Access                                             Med Practice: Ease of scheduling                       100                                   Estblished

101             Med Practice: Access                                              Med Practice: CP Efforts                                      100                                   Estblished

102            Med Practice: Care Provider                                 Med Practice: CP Concern                                   100                                    New

102             Med Practice: Care Provider                                 Med Practice: CP Explanation                            100                                    New

102             Med Practice: Care Provider                                 Med Practice: CP Discuss treatments              100                                    New

102             Med Practice: Care Provider                                 Med Practice: CP Efforts                                      100                                    New

102             Med Practice: Access                                            Med Practice: Ease of contacting                         100                                  New

102             Med Practice: Access                                             Med Practice: Ease of scheduling                       100                                  New

102            Med Practice: Access                                              Med Practice: CP Efforts                                      100                                  New

[PaigeMiller]

Which two groups do you mean?

Statistically different based on what statistical test?

[robertrao]

THE TWO groups are : under Appointment_Type variable.
Established and NEW.

I am guessing it will be a correlation procedure. I might be wrong!

Thank you

[PaigeMiller]

Correlation not usually used to show "statistically different".

Can you describe in words what comparison of the data you are thinking about? Please provide enough detail so we can follow along.

[Ksharp]

Since your question is about STAT, better post it at stat forum:
https://communities.sas.com/t5/Statistical-Procedures/bd-p/statistical_procedures

Let more statistical experts to see it.
@StatDave @lvm @SteveDenham @jiltao ........

[SteveDenham]

There are simple ways, and more complete ways. For the simple way try:

proc ttest data=your_data;
by question;
class appointment_type;
var score;
run;

However, it appears that your data may be clustered by ID, with several questions for each. If there is any chance that the response within an ID is correlated, then perhaps a generalized estimating equation approach would be useful. Here is some example code:

proc gee data=your_data;
class appointment_type question ID;
model score = appointment_type question;
repeated subject=ID/ type=exch covb corrb;
run;

You may want to include an interaction term, if you want to compare mean scores for each question. That would lead to this code:

proc gee data=your_data;
class appointment_type question ID;
model score = appointment_type question appointment_type*question;
repeated subject=ID/ type=exch covb corrb;
lsmeans appointment_type*question;
slice appointment_type*question / diff sliceby=question;
run;

Hope some of this helps.

SteveDenham

[robertrao]

Hello Steve, Thank you for the code and explanation.

We noticed that there is a good rating score given for the questions in the Established category compared to the NEW and we did a TTEST of the averge score per ID between the two groups and found to be statistically significnt.

Now we are trying to find if the Established versus NEW groups had any correlation for the questions they answered.

For example : IF Med Practice: CP Discuss treatments question score was significantly different between the two groups.....or if other question did etc. etc.

Would PROC GEE still a good procedure for this kind of analysis?

Thank you, again

[SteveDenham]

Yes, it would. It would enable you to account for correlation within ID as well as the correlation/covariance across ID's.

SteveDenham

[robertrao]

ID is the patient ID and each patient takes the survey and gives the score for each question in the domain(s).

When we look for any correlation like in our study trying to find which question did the Established favor more ( since we know Established had high Mean scores), do we need to use ID in the PROC GEE?

Also would you share a document to interpret the results ?

This is the result after using PROC GEE. Does this say anything ? If there was a difference for any of the questions each group answered?

Thank you

[SteveDenham]

I see the solution vector, but the sliced least squares mean differences are what would address your research question.

SteveDenham