UGA MARK 4000 Exam 2

primary data

information collected specifically for the purpose of the investigation at hand

secondary data

information not gathered for the immediate study at hand but for some other purpose

why secondary data first?

- significant advantage in time and money

- with secondary data, someone else has already paid for data collection

- even if there is a charge for using the data, the costs are significantly lower than they would be if you collected the information yourself

- sometimes secondary data might be sufficient

types of secondary data: internal

sales data (volume/dollar transaction, products/services sold, discount applied); sales person or agent responsible for sale; location of customer transaction (loyalty card)

types of secondary data: external, published

government (census); newspapers/periodicals; annual reports; market research/ survey by third party

types of secondary data: external, commercial

purchase panels, scanner data, geo-demography, ad exposure data, behavioral measurement

types of primary data

- Demographic/ socioeconomic characteristics

- Personality/lifestyle characteristics

- Attitudes

- Awareness/knowledge

- Intentions

- Motivation

- Behavior

primary data can be collected via:

- communication: oral or written

- observation: recording of behavior

primary data can be in the form of:

- numerical values (quantitative)

- non-numerical values (qualitative)

main advantage of COMMUNICATION over OBSERVATION:

elicitations of unobservable (opinions, beliefs)

main advantage of NUMERICAL over NON-NUMERICAL form:

enable concrete measurement that that represent information that enables summarization/analysis

nominal scale

- objects are either identical or different

- classification/ identification (ex. jersey number, social security card)

- %, mode

- dichotomous (yes or no) scale

ordinal scale

- objects are greater or smaller

- ordering of objects preference (ex. rank)

- %, mode, median

- rank order

interval scale

- intervals between adjacent numbers are meaningful

- attitude scales

- %, mode, median, mean, std. deviation

- Likert scale, semantic differential

Ratio scale

- meaningful zero; comparison of absolute magnitudes is possible

- sales, weight, age

- almost all kinds of stats.

- fill in the blank (open-ended)

mean

the sum of the values for all of the observations of a variable divided by the number of observations

median

the observation for which 50% of the observations fall

mode

the value that occurs most frequently

Why use other scales than ratio (most versatile)?

1) attribute/characteristics of interest dictate the type of scale used (attribute must possess a natural/absolute zero that reflects "absence" of the attribute (ex. height weight)

- not all attributes have absolute zero property (sex: M/F, attitudes)

2) ratio scale may be too intrusive in questionnaire setting (participants may not be comfortable providing age/income

ex. "What is your age?" v.s. "Which best describes your age group?"

types of self reports

- self reports: people are asked for their beliefs of feelings

1) itemized rating scales

2) graphic rating scales

3) comparative rating scales

summated rating scale (Likert)

subjects indicate their degree of agreement or disagreement with a series of statements that are favorable (F) or unfavorable (UF) to the construct of interest

- possible to attain summated or total score across items

- score is interpreted *always* by comparing with another sample, object, or time period to be meaningful

semantic differential scale (bipolar)

can measure respondent attitudes towards ideas, concepts, items, people, and events. (Sometimes referred to as an attitudinal study); scale between two bipolar adjectives, such as "Happy-Sad," "Creamy-Chalky," or "Bright-Dark."

graphic ratings scale

a scale in which individuals indicate their ratings of an attribute typically by placing a check or slider at the appropriate point on a line that runs from one extreme of the attribute to the other

- more granular/infinite level of items to choose from

- drawbacks: often small sample for each "Choice", measurement is often a pain in paper/pen survey- now less of a concern

constant sum scale

a comparative-ratings scale in which an individual divides a given sum among two or more attributes on some basis, such as importance or favorability

- drawback: might be hard for participants to answer

Which scale to use?

- interval scale will most likely be the choice for attitude (Likert is most popular)

- nominal scale for background/demographic information

- number of scale positions for interval scale (5-9, even or odd number?); neutral position appropriate? Odd number is prefered

- inclusion of "dont know" or "not applicable" (should include if sizable number of participants are new to the topic/havent encountered)

if your participants are busy, distracted, bored, unmotivated, or tired, use:

likert vs constant sums, 5-6 interval vs 8-9

if your participants are concerned about privacy, embarrassment, use:

nominal (yes or no) vs ratio (open ended)

General structure of questionnaire

1) introduction

2) Questions (content, format)

3) Question wording

4) determine question sequence

Questionnaire: Introduction

- information on the surveyor/statement on the purpose of the survey

- how the respondents were selected/incentive if at all

- screening questions- screen out respondents based on objectives of the survey (e.g. tourist vs. resident/student)

2) Questions

- warm up questions- complicated/difficult to answer questions (interval scales)

- classification questions (nominal scale) eg. age, gender, income

the process of questionnaire design

1) planning what to measure (Research objective, research issues, what to be asked)

2) formatting the questionnaire (content and format)

3) question wording (question wording, evaluate each research question)

4) sequencing and layout (layout, group in each subtopic)

5) pretesting and correcting problems (check for sense, check for error, pretest corrections)

Step 2: Content

- how many questions? Capture the needed data in as few questions as possible

- is the question answerable? Do they know or are they unwilling?

- may not answer if its too much effort or sensitive

Step 2: Format

includes open ended questions or close ended questions

open ended questions

versatility, can be either factual or exploratory (difficult to code, analyze, or interpret the answers

close ended questions

easy to answer, code, analyze, and interpret; response categories must be exhaustive (5-7 categories); response categories must be mutually exclusive, except in special cases where more than one answer is acceptable; include "other" option when necessary

multichotomous questions

a closed-ended question in a marketing research questionnaire in which a respondent must choose one response from two or more possible alternatives

order of response categories

- Can affect responses

- To prevent order bias, place the average or expected

response at various positions in the sequence of

categories

leading question

a question that implies that one answer would be better than another

ex: Why do you like Wendys fresh meat compared to those of competitors?

loaded question

A question that is biased because it contains a built-in assumption

ex: Have you stopped visiting restaurants in downtown Athens?

Double barreled question

a single survey question that actually asks two questions but allows only one answer

ex: Do you eat Wendys hamburgers and chilli?

overstated question

places undue emphasis on some aspect of the topic

ex: How much would you spend on a pair of sunglasses that will protect your eyes from the suns harmful UV rays, which are known to cause blindness?

non-exhaustive question

does not have enough choices to answer

ex: Where do you live? [] home [] dorm

non-mutually exclusive answers

what is your age? under 20, 20-40, 40 and over

ambiguous question

a question containing a concept that is not defined clearly

ex: Do you eat at fast food restaurants regularly?

unanswerable question

What was the occasion for eating your first hamburger?

words to avoid

the words in a questionnaire should have a single meaning that is known to the respondents

ex: ambiguous words- usually, normally, frequently, often, regularly, occasionally, sometimes

ten words to avoid in question development

all, always, any, anybody, best, ever, every, most never, worst

implicit alternatives

an alternative that is not explicitly expressed

ex: Do you like to fly when traveling a short distance? v.s. Do you like to fly when traveling a short distance or would you rather drive?

implicit assumptions

assumptions that are not stated in the question

ex: Are you in favor of a balanced budget? v.s. Are you in favor of a balanced budget if it would result in an increase in the personal income tax?

2 Do's of question wording

1) simple/brief sentence: grammatically simple sentence with 20 words or less

2) focused/clear: everyone should interpret the question in the same manner

Step 4: determine question sequence

- use simple & interesting open ended questions

- use the funnel approach: start broad and progressively narrow down the scope

- design branch questions with care

- ask for classification questions last (basic info first- attitudes, intentions, perceptions; then classification- demographics)

- place difficult or sensitive questions last

tips for handling sensitive questions

- guarantee respondents that their answers will be completely anonymous

- put any sensitive questions near the end of the questionnaire

- use a counter biasing statement that indicates the behavior or attitude in question is not unusual

- phrase the question in terms of other people & how they might feel or act

- ask for general answers, rather than specific ones

step 5: pretest questionnaire

- vital, most inexpensive insurance you can buy to ensure success of the questionnaire and project

- data collection NEVER begin until you have pre-tested and revised

- last chance for researcher to make sure the data collection is working properly

2 recommended pretests

1) face to face or personal interview pretest

2) actual pretest using chosen method of administration

recruiting respondents

- good cover letters and scripts are not written in a hurry

- the usual things to include are: who you are, why you are contacting them, the request for help, how long it will take, promise of anonymity or confidentiality, any incentives

How to perform reverse scaling

1. Reverse the score the negative items. If the responses were on a 5-point Likert scale, the positive items are scored 0, 1, 2, 3, 4. Then the negative items should be rescored as 4, 3, 2, 1, 0.

- ONLY do this for the negative items & find new sum

2. Use mean value or median according to the data. And use the new scale in which your have changed the values.