Analyzing Survey Data

Home > Statistics Every Writer Should Know > The Stats Board > Discusssion

Analyzing Survey Data
Message posted by Tony Vallin (via 4.34.180.42) on January 28, 2002 at 5:12 PM (ET)

Recently, my wife’s school was presented the results of a survey that was conducted regarding the effectiveness of their school’s principal. The survey had 2 discrete sections. Section 1 was in the form of a question, pertaining to a specific practice of the Principal, with response values being 1=Unsatisfactory through 7 = Outstanding. A response of 4 = Acceptable. When the results of this section were presented by the school boards guru, he simply averaged the responses. For example, if 10 persons responded 1=Unsatisfactory and 10 persons responded 7=Outstanding, the result was calculated as ((10*1) + (10*7))/20=4. My wife argued, and I concur, that this was an invalid methodology for analyzing this section of the survey since all results would fall at or around the average of 4. The acceptable limits in this section of the survey were between 3.5 – 4.5. And of course, the actual result was 4.5.
Section 2 was worst. Similar in format to Section 1, the response values in Section 2 were 1=Strongly Disagree 7 = Strongly Agree with 4 = Neither Agree nor Disagree. Using the same averaging technique the actual result calculated to 4.8. Again the acceptable limits for this section were between 3.5 – 4.5.
I would appreciate any input into A) the validity of using this technique for analyzing this instrument and B) if not the right technique, what would have been more appropriate.

TIA

READERS RESPOND:
(In chronological order. Most recent at the bottom.)

Re: Analyzing Survey Data
Message posted by JG (via 128.8.22.41) on January 29, 2002 at 5:12 AM (ET)

See the statistics for social science literature on the construction of Likert scales, etc.
Otherwise, the thing to do is to construct frequence distributions and do correlation analysis. This should clarify what the data really shows. Does the population divede into those who like the principal, those who do not like him, and those who are indifferent?, etc.

Re: Analyzing Survey Data
Message posted by Phil (via 66.32.156.26) on February 12, 2002 at 10:25 AM (ET)

Calculating means and standard deviations of Likert scale data is not considered appropriate analysis by all survey experts. In my doctoral dissertation I used survey data and my chair wouldn't let me include any standard deviations because he felt so strongly about it. In your case using the mean masks the results and is not doing what is intended. No one is going to pay attention to the standard deviation anyway.

Sometimes it is more useful for analysis to add the "High" scales together and get a percentage that lends itself to correlation analysis, comparison, SPC, etc. For example, suppose Very High = 20% and High = 30%. The total of 50% would be your statistic.

Your $5 contribution helps cover part the $500 annual cost of keeping this site online.