Descriptive Data Analysis Analyzing Survey Data Research Methods

Descriptive Data Analysis: Analyzing Survey Data Research Methods for Public Administrators Dr. Gail Johnson Dr. G. Johnson, www. Research. Demystified. org 1

At Long Last: Analyzing Surveys: Using Percent Distributions Key Topics: Handling “Exits” Analyzing 5 -point scales Extreme analysis Handling the middle categories Dr. G. Johnson, www. Research. Demystified. org 2

Guidelines for “Exits” Exit options on scales: “don’t know” “not applicable” “no opinion” Handling “Exits”: no firm rules Dr. G. Johnson, www. Research. Demystified. org 3

Guidelines for “Exits” Key decision: leave in or exclude from analysis? If the number of people taking an exit is very small, there will be little difference in the percent distribution so the researchers only need to inform the reader whether they are included or excluded Dr. G. Johnson, www. Research. Demystified. org 4

Handling “Exits” 1. If there is a high percent of exits, does it make sense? If it does, exclude from the analysis For example, if you ask people how satisfied or dissatisfied they are with the zoning office, a high percent might say “not applicable. ” 2. That might make sense because probably only a small proportion of citizens contact the zoning office in any given year. 3. The “not applicable” responses could be excluded, and the analysis should focus on the percent distribution who actually answered in the satisfied—dissatisfied scale. 1. Dr. G. Johnson, www. Research. Demystified. org 5

Handling “Exits” 1. If, on the other hand, there is a high proportion of people taking the exits for reasons that are not clear, the researchers might want to spend some time to figure that out. 1. It could be a sign that there was a problem with the question. 2. My best advice—as a general rule— would be to exclude the exits from the analysis, and focus on only those who answered within the scale. Dr. G. Johnson, www. Research. Demystified. org 6

Handling “Exits” 1. Make sure the analysis is anchored in terms of the number of people who answered within the scale. 2. “Of the 300 people who indicated their level of satisfaction, 65% reported being very or somewhat satisfied. ” Dr. G. Johnson, www. Research. Demystified. org 7

Analyzing Survey Scales Decision rules for 5 -point scales Very Satisfied Somewhat satisfied Neither Somewhat dissatisfied Very dissatisfied Dr. G. Johnson, www. Research. Demystified. org 8

Analyzing Survey Scales 1. Decision rules for 5 -point scales. 1. I first look at where a majority (50% or more) answered either on the positive or the negative side of the scales. 1. 2. Very satisfied and somewhat satisfied. Very dissatisfied and somewhat dissatisfied. It allows me summarize the set of questions: 1. “Of the 5 questions about the city services, a majority reported being very satisfied or somewhat satisfied with library services, recreation department and the fire department. ” Dr. G. Johnson, www. Research. Demystified. org 9

Analyzing Survey Scales: “Extreme Analysis” 1. If a majority answered positively or negatively on many questions, I then do an “extreme analysis”—that is, I look for questions where a majority answered “very satisfied” or “very dissatisfied. ” 2. It is unusual to get a majority at the extreme end of scale, so that information shows the greatest intensity of feeling about the issue. Dr. G. Johnson, www. Research. Demystified. org 10

Exercise: What’s The Story About Faculty Performance? In general, how satisfied or dissatisfied are you with various aspects of faculty performance? Very Satisfied Generally Satisfied As Satisfied as Dissatisfied or Neither Generally Dissatisfied Very Dissatisfied 54% 40% 5% 1% 0% b) Ability of faculty to relate concepts to real world settings 37 51 12 1 0 C Willingness of faculty to meet student needs 24 39 12 14 10 d) Availability of faculty outside of class 22 38 14 16 10 a) Level of faculty knowledge Dr. G. Johnson, www. Research. Demystified. org 11

Exercise: What’s The Story? What are students most satisfied with? What are students least satisfied with? What’s the story the data tells? Dr. G. Johnson, www. Research. Demystified. org 12

Analyzing Survey Scales 1. A majority was very satisfied or generally satisfied with all four aspects of faculty performance. 1. However, a majority (54%) were very satisfied with faculty knowledge. 2. Part of analyzing the data requires identifying the key elements of the story. 1. I would highlight faculty knowledge as being a strong aspect of faculty performance according to those completing the survey. Dr. G. Johnson, www. Research. Demystified. org 13

Handling the Middle Category As you know, I like uneven scales (scales with 5 or 7 choices). For example, Likert scales have a neutral middle. Neither agree or disagree. Combining the “neither” category with the agree side or the disagree side clearly does not make any sense. Dr. G. Johnson, www. Research. Demystified. org 14

5 -point Scale Without a Neutral Middle (one-way scales) 1. 2. 3. 4. 5. ___ Almost Never/Never ___ Seldom ___ Occasionally ___ Usually ___ Almost Always/Always Note: a graduated one-way scale with soft-ends Dr. G. Johnson, www. Research. Demystified. org 15

Handling the Middle Category Another common one-way scale Very great extent Great extent Moderate extent Some extent Little of no extent Dr. G. Johnson, www. Research. Demystified. org 16

Handling the Middle Category in One-way Scales It is possible to combine categories when using a scale without a neutral middle, It could be honestly reported that 90 percent of the respondents rated the city services as helpful to at least some extent. However, the researcher is rolling 4 of the 5 possible categories together—and raises a question of whether the researcher is trying to hide the detail. Maybe very few respondents provide favorable ratings. Dr. G. Johnson, www. Research. Demystified. org 17

Handling the Middle Category in One-way Scales 1. My advice: the middle category stands alone, even though it is not neutral. 2. It is too easy to distort the data by combining the middle category with either side of the scale to make it say what the researcher wants. Dr. G. Johnson, www. Research. Demystified. org 18

Handling the Middle Category in Goldilocks Scales Does the program offer too many, too few are just about the right number of electives: Much too many Somewhat too many Just about right Somewhat too few Much too few Dr. G. Johnson, www. Research. Demystified. org 19

Handling the Middle Category in Goldilocks Scales The middle category is not neutral: it contains important information. In fact, the program director would be hoping that most people say “just about right” It too stands alone! Dr. G. Johnson, www. Research. Demystified. org 20

Handling the Middle Category My rule: the middle always stands alone. Let me repeat: the middle stands alone. Dr. G. Johnson, www. Research. Demystified. org 21

Why Not Have Yes or No Scales? 1. Or even scales? 2. I prefer 5 -point scales because it gives me the option of looking at the extreme ends of the scales. 3. If I only ask people whether they are satisfied or dissatisfied, I do not know how many are really very satisfied as compared to somewhat satisfied. 4. The difference in intensity might be important. Dr. G. Johnson, www. Research. Demystified. org 22

Why Not Have Yes or No Scales? I still have the option of collapsing the 2 positive together and 2 negative sides of the scale together to simply reporting. The middle category gives a place for people to go if they really do not have an opinion one way or the other. Some people do not like to say they “don’t know”, so they find a middle category. Dr. G. Johnson, www. Research. Demystified. org 23

Key Questions About Survey Results Make sure you know the decision rules: did the researchers exclude the people answering “don’t know”? Did they do an extreme analysis? How many categories did they roll together to get their results? Did they maintain the neutral middle? Dr. G. Johnson, www. Research. Demystified. org 24

Final Words on Analyzing Survey Data 1. Be mindful of words like “most people”—do they mean the majority (50%+) or do they mean a plurality (the greatest proportion but less than a majority). 2. When working with survey data, round percentages to the nearest whole number (. 5 and up, round up, less than. 5, round down). 1. Decimal points can give a false sense of precision and make it harder for people to remember the numbers. Dr. G. Johnson, www. Research. Demystified. org 25

Creative Commons This powerpoint is meant to be used and shared with attribution Please provide feedback If you make changes, please share freely and send me a copy of changes: Johnsong 62@gmail. com Visit www. creativecommons. org for more information Dr. G. Johnson, www. Research. Demystified. org 26