ASA Connect

Hello,

I'm dealing with a situation where a client would like to use a questionnaire in which each subject is ask to list some behaviors and rate them in importance.
The complication is that is an open-ended questionnaire , so that each subject can come up with a different list.  Even if the superset of responses is a limited list, there will be differing subsets for each subject (e.g., 'A,B,D', 'B,C,E', ....).

I had thought of finding the superset, and for each subject, setting their importance rating to 0 if they did not list an item.  For example, the subject above would have ratings for A, B, C,D, E with their ratings for C, E set to 0.

Does anybody have any thoughts on this, or references to people dealing with this situation before?

Sincerely,

Barry DeCicco

------------------------------
Barry DeCicco
------------------------------

In a presentation I gave several years ago to some Ford executives and others, we had the same issue. We found that the open ended questions fell into 5 categories us an "other" group. We put a "1" when there was a complaint in that category and a "0" otherwise.  From that, we found that 4 of the 6 columns and 2 "offices" were all the drivers for bad reviews for the company we worked for. 

We used a CART ensemble to find those important drivers. Each ensemble had a different set of data in each partition. So, CART1 had the total data partitioned one way. CART2 had the original data partitioned a second way, etc. Then we took the most common "important factors" and reported them. That allowed us to eliminate most spurious correlations. We only had 600ish responses. So, spurious relations were a big issue.

------------------------------
Andrew Ekstrom

Statistician, Chemist, HPC Abuser;-)
------------------------------

Original Message:
Sent: 03-19-2021 14:31
From: Barry DeCicco
Subject: Regression with Differing Sets of Predictors

Hello,

I'm dealing with a situation where a client would like to use a questionnaire in which each subject is ask to list some behaviors and rate them in importance.
The complication is that is an open-ended questionnaire , so that each subject can come up with a different list.  Even if the superset of responses is a limited list, there will be differing subsets for each subject (e.g., 'A,B,D', 'B,C,E', ....).

I had thought of finding the superset, and for each subject, setting their importance rating to 0 if they did not list an item.  For example, the subject above would have ratings for A, B, C,D, E with their ratings for C, E set to 0.

Does anybody have any thoughts on this, or references to people dealing with this situation before?

Sincerely,

Barry DeCicco

------------------------------
Barry DeCicco
------------------------------

Does ANALYSING ORDINAL DATA, SURVEYS, COUNT DATA:USING R TO ANSWER SOCIOLOGICAL QUESTIONS provide any useful insights?
Bill

------------------------------
Bill Harris
Data & Analytics Consultant
Snohomish County PUD
------------------------------

Original Message:
Sent: 03-19-2021 14:31
From: Barry DeCicco
Subject: Regression with Differing Sets of Predictors

Hello,

I'm dealing with a situation where a client would like to use a questionnaire in which each subject is ask to list some behaviors and rate them in importance.
The complication is that is an open-ended questionnaire , so that each subject can come up with a different list.  Even if the superset of responses is a limited list, there will be differing subsets for each subject (e.g., 'A,B,D', 'B,C,E', ....).

I had thought of finding the superset, and for each subject, setting their importance rating to 0 if they did not list an item.  For example, the subject above would have ratings for A, B, C,D, E with their ratings for C, E set to 0.

Does anybody have any thoughts on this, or references to people dealing with this situation before?

Sincerely,

Barry DeCicco

------------------------------
Barry DeCicco
------------------------------

Thank you, Bill!  I'm reviewing that.

------------------------------
Barry DeCicco
------------------------------

Original Message:
Sent: 03-22-2021 20:00
From: Bill Harris
Subject: Regression with Differing Sets of Predictors

Does ANALYSING ORDINAL DATA, SURVEYS, COUNT DATA:USING R TO ANSWER SOCIOLOGICAL QUESTIONS provide any useful insights?
Bill

------------------------------
Bill Harris
Data & Analytics Consultant
Snohomish County PUD

Original Message:
Sent: 03-19-2021 14:31
From: Barry DeCicco
Subject: Regression with Differing Sets of Predictors

Hello,

I'm dealing with a situation where a client would like to use a questionnaire in which each subject is ask to list some behaviors and rate them in importance.
The complication is that is an open-ended questionnaire , so that each subject can come up with a different list.  Even if the superset of responses is a limited list, there will be differing subsets for each subject (e.g., 'A,B,D', 'B,C,E', ....).

I had thought of finding the superset, and for each subject, setting their importance rating to 0 if they did not list an item.  For example, the subject above would have ratings for A, B, C,D, E with their ratings for C, E set to 0.

Does anybody have any thoughts on this, or references to people dealing with this situation before?

Sincerely,

Barry DeCicco

------------------------------
Barry DeCicco
------------------------------