Return-Path: <nifl-health@literacy.nifl.gov> Received: from literacy (localhost [127.0.0.1]) by literacy.nifl.gov (8.10.2/8.10.2) with SMTP id fBHGHl010314; Mon, 17 Dec 2001 11:17:48 -0500 (EST) Date: Mon, 17 Dec 2001 11:17:48 -0500 (EST) Message-Id: <3C1E1AB2.665DC65C@emory.edu> Errors-To: listowner@literacy.nifl.gov Reply-To: nifl-health@literacy.nifl.gov Originator: nifl-health@literacy.nifl.gov Sender: nifl-health@literacy.nifl.gov Precedence: bulk From: Holly Avey <havey@emory.edu> To: Multiple recipients of list <nifl-health@literacy.nifl.gov> Subject: [NIFL-HEALTH:3497] likert scales X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset=us-ascii X-Mailer: Mozilla 4.5 [en] (Win95; I) Status: O Content-Length: 5289 Lines: 114 A few weeks ago I posted a question about using likert scales for low-literacy readers. Specifically, I was asking about the answers to these questions violating normal distribution assumptions. I had a few requests to post the answers to the listserve, so here they are below (I have also copied my original question at the bottom). The main themes for the answers seem to be: 1) Likert scales tend to have answers that cluster on either end anyway, and that's okay 2) Write out the answers whenever possible and position them vertically 3) Whether or not you use labels along with the numbers, how you ask people to indicate their answers, and how many points there are on the scale can all affect the answers you receive 4) Avoid the use of likert scales 5) If you have the opportunity in an education class, use the scale as an activity with people standing in different parts of the room to correspond with different answers, followed by a discussion and opportunity to change positions Thanks to all who responded!! -- Holly P.S. I'm afraid I was unable to open a message from Carol Barks Bani regarding this topic, so this summary does not include her response. *************** The normal distrubution is made-up. Having answers cluster at either end of a likert scale is completely OK. There are a lot of issues around the "normal" distrubtion. Others more test savvy can probably fill in with more info. Example: a lot of consternation at Harvard about grade inflation. Who knows, it is possible for all students to get A's, it doesn't have to be "grade inflation." *************** There is a strong tendency for Likert scale responses to cluster at the ends anyway. I don't suggest using Likert scales. Even though it is more work, write out the choices vertically: 1. I like school. a. Always b. almost always c. usually etc. **************** I remember hearing Rima Rudd (Harvard literacy expert, and excellent speaker if you ever have the chance). She said that materials for low literacy adults are often longer in order to be clear. Writing out responses as Kathleen suggests is a great idea **************** When using Likert-type scales with low-literacy readers, do you label the end-points of the scale (e.g., "a lot, very true" at one end and "not at all, false" at the other)? Do you label the mid-point? Or do you simply give instructions that say something like "1=least, 7=most"? This may well have something to do with the various successes people are reporting. Also, are you using 5-point or 7-point scales? Asking people to circle a spot on a scale, circle a number, or write a number in a blank? All these may also be relevant. **************** With many low literacy readers, it isn't subtle variations that are the problem, but the fact that a likert scale was used to begin with. Also, rarely will a low-literacy reader make the effort to read the directions. *************** We have used the likert scale in a different way for low-literacy students in adult education classes. We post signs on either side of a wall that say "I strongly agree" and "I strongly disagree". In the middle is a sign that says "not sure". (Or you could use "always/never" or whatever other responses). We read statements (for example: "All women over 40 should get a mammogram every year.") and the people go and stand along the wall in the place that corresponds with how they would answer, either under a sign or in between somewhere. Then we have volunteers say why they are standing where they are, and a discussion begins. Some people choose to move to a different spot in response to the discussion. Obviously this would not work for research or in a medical setting, or other of the ways you use these scales. But in an education class, where the goal is discussion and learning, rather than collecting statistics, it is a fun and engaging way to use the likert scale. ***************** My original question: I recently tried to analyze some data I have been collecting from a low-literacy population, and I ran across an issue I suspect others already know about and have researched. I attempted to write the survey questions for my pre-test at an appropriate readability level, based on general assessments of our patient population a few years ago. Many of the questions featured likert scales for answer options. When I began to analyze the data, I discovered that it violated normal distribution assumptions. Specifically, the data seemed to cluster at the extreme answer options. My suspicion is that people with low literacy may be more likely to choose extreme options on likert scales, possibly because the subtle variations between the choices get a little confusing. Does anyone know if this phenomena has been tested empirically? If so, could you provide me with some references? Any suggestions on how to address this problem? (Maybe only offer three choices instead of five, or throw out all likert scales completely?) It does pose quite a problem when trying to use standardized instruments and compare results to other population groups. . . ****************** Holly Avey, MPH Health Educator, Office of Health Promotion Grady Health System, Atlanta, GA Ph: 404-616-7561, Fax: 404-880-9464, E-mail: havey@emory.edu
This archive was generated by hypermail 2b30 : Fri Jan 18 2002 - 11:28:45 EST