" "whether certain question responses tended towards one of the extremes" .caught my eye. Given that you have a Likert scale (not uncommon in "patient reported outcomes" PROs'), if you see most responses at the highest value or the lowest value, then the scale may have what the PRO folks call a "ceiling" or a "floor" effect. An example for analysis of a PRO with ceiling effects here The ceiling effects of patient reported outcome measures for total knee arthroplasty - ScienceDirect
------------------------------
Chris Barker, Ph.D.
Past Chair
Statistical Consulting Section
Consultant and
Adjunct Associate Professor of Biostatistics
www.barkerstats.com---
"In composition you have all the time you want to decide what to say in 15 seconds, in improvisation you have 15 seconds."
-Steve Lacy
------------------------------
Original Message:
Sent: 06-19-2024 14:51
From: Andrea Mack
Subject: Sentiment Survey Analsyis
Tom,
I appreciate your response and can envision what you're describing. The graphical display you described will do a better job of showing the distribution of responses and whether certain question responses tended towards one of the extremes.
Thank you!
Andrea
------------------------------
Andrea Mack
Statistician
Original Message:
Sent: 06-17-2024 13:01
From: Thomas Sandry
Subject: Sentiment Survey Analsyis
Andrea,
The only approach that springs to mind is to graphically display the results in a statistical control strip chart format, with control limits set at say between 1 and 2, and also between 4 and 5, on the Likert scale. For convenience in printing you could use a vertical orientation of the chart, where each horizontal line represents a question, and the leftmost portion of each horizontal line can serve as the response descriptor. The vertical columns which the response characters will naturally fall in can have headings 1 - 5 at the top of page and a vertical bar symbol can be used to designate the control limits by placing them between columns 1 and 2 and also between columns 4 and 5. Ordinary characters such as x or * can be used to designate the individual response Likert values on each line since they will all line up in vertical columns, since the responses are integers, not continuous values. If this seems like a useful visualization I can try to draw up an example using simulated data for consideration. This simple presentation strategy may be available in code, but it's also very simple to implement on a typewritten page, so maybe no one has bothered to program it.
The resulting 16 line graphic will draw attention to those responses ( lines ) which fall outside, left and right, of the vertical control limit format lines and if it was convenient, the red and green colors could be used for emphasis. Or the lines with extreme responses could simply be highlighted in the appropriate color.
Hope this helps,
Tom
Thomas D. Sandry, PhD
Industrial Statistical Consultant, Retired
------------------------------
Thomas Sandry
Original Message:
Sent: 06-17-2024 11:47
From: Andrea Mack
Subject: Sentiment Survey Analsyis
Hello Section Members,
I'm looking for some guidance from any of our members that are well versed in suvey analysis. An organization completed a sentiment survey of a few (16) targeted employee's feelings about core capabilities, staffing, and infrastructure readiness in several areas. The feelings were rated on a 1-5 Likert scale, ranging from "Strongly Disagree" to "Strongly Agree". Results have previously been summarized two ways: (1) using the continuum of red-yellow-green indicators for the average response for each question, where red is strongly disagree, yellow is neutral, and green is strongly agree and (2) colors indicative of the results from a hypothesis test, under the null hypothesis of the average response by question being "Neutral". Does anyone in the community have advice/guidance on how else to analyze such sentiment survey data, so that questions with extreme responses stand out?
Thanks,
Andrea
------------------------------
Andrea Mack
Statistician
------------------------------