Library Surveys How to approach them Jane M' Horgan, Dublin City University

1
Library SurveysHow to approach themJane M.
Horgan, Dublin City University
ANTLC ACADEMIC AND NATIONAL LIBRARY TRAINING
CO_OPERATIVE
Quality Measures The Irish Academic Library
Context Hosted by the Library, National
University of Ireland Maynooth Mon 22nd November,
2004 Glenroyal Hotel, Maynooth
2
Overview

• What is a survey?
• Steps in a survey
• Selecting a representative samplè
• Pitfalls

3
What is a survey?

• A method of gathering information from a
  subgroup/sample to infer to the whole
  group/population.
• Examples
• Evaluation of the smoking ban
• Political opinion polls
• ANLTC survey to identify future training needs of
  members
• Library Survey NUIM 2003 to identify how service
  can be improved

4
Why take a sample?

• Time
• Cost
• Accuracy

5
A

A Representative Sample
Population
Sample
?
Purpose of sample To infer to population Sample
must be chosen so that it is representative of
the population
6
Library Survey NUIM 2003Objective to identify
how service can be improvedPopulation Full-time
academic staff and students
What is a library survey?

• Academic Staff
• Arts
• Science
• Theology

• Students
• Research postgraduates
• Taught postgraduate
• Undergraduate

A stratified random sample design
7
Surveys differ with respect to
Types of Surveys
Purpose/Subject matter

• Coverage
• Sample
• Census

• Source of information
• Official statistics
• Observational studies
• Mail questionnaire
• Telephone interview
• Face-to-face interview
• Internet questionnaire

8
Steps in a survey

• State objectives
• Define population
• Decide on data required to fulfill your
  objectives
• Decide on sample design
• Decide on method of measurement
• Carry out pilot study
• Revise questionnaire and sample design
• Collect data
• Analysis
• Final report

9
What can go wrong?

• Sample not representative of population
• Questionnaire design
• No pilot
• Non-response
• Sample too large for budget and time constraints

10
What can go wrong?
1. Unrepresentative Sample
Population
Sample
?
Purpose of sample To infer to population
Sample must be chosen so that it is
representative of the population
11
A

A Representative Sample
Population
Sample
?
12
A representative sample
Library of Books
Sample
? infer
13
How to select a representative sample
Example A librarian has an catalogue of 500
books, and wants to select 50 to estimate the
usage rate

• One number is selected randomly between 1 and 10,
  7 say.
• The books corresponding to the numbers 7, 17,
  27, etc are selected.
• This is called a Systematic Random Sample

• A set of 50 numbers between 1 and 500 is chosen
  randomly.
• The books that correspond to these numbers are
  selected into the sample
• This is called a Simple Random Sample

14
Random (Representative) Samples

• A random sample is a sample chosen so that every
  item in the population has a chance (probability)
  of selection into the sample

• To select a random sample you will need a list
  of all items in the population.
• List of all staff at university
• Database of categories of students
• Database of all books in the library
• List of members of ANLTC

If list is inadequate (gaps), sample may not be
representative
15
Stratified random sampling

• Stratify the population by
• Gender
• Age
• Stratify student/academic by
• Faculty
• Level
• Stratify library
• Classification type
• Subject

Take a simple random sample from each
group/stratum
16
Other methods of random sampling

• Systematic sampling Subjects are selected fro
  list/frame after a random start
• Cluster sampling Population divided into groups
  and a random sample of groups is taken.
• Example Sample student population by classes

17
Non-random sampling

• Judgement sampling
• Volunteer sampling
• Haphazard sampling
• Availability sampling
• Quota sampling

18
A

A Representative Sample
Population
Sample
?
Purpose of sample To infer to population Sample
must be chosen so that it is representative of
the population
19
Bad questions
2. Questionnaire Design
What can go wrong?

• Do you visit the library regularly?
• How would you rate the convenience and quality of
  the library service? (v.good, good, fair, poor,
  v.poor)
• What is you income?
• Did you ever steal a book from the library?
• Tick your age category? (15-20, 20-30, 30-40,
  gt40)
• What percentage of your week do you spend in the
  library?
• Rank in order of importance the following nine
  factors.
• Do you use CD roms? (Yes, No)
• Do you find the library noisy? (Yes, No)
• How many times did you visit the library last
  year?

20
Questionnaire Construction

• KISS Keep it short, simple, specific
• Logical question progression
• Flow more important than grammatical correctness
• Key questions first, sensitive last
• Keep allied questions together
• Be reasonable when requesting forecasts
• Mix fully structured and open-ended questions

21
Some Advice about Questionnaire Design

• Write down objectives of your survey
• Have a purpose for every question
• Prepare tables before you field your
  questionnaire
• Pilot test your questionnaire (more than once may
  be necessary)

Recommended Reading Sudman Bradburn Asking
Questions, San Francisco Jossey Bass, 1982
22
Pilot Study
What can go wrong?
3.No time to do pilot study

• A pilot study is a pretest of your methods.
• It is carried out on a small group of typical
  people as a dress rehearsal.
• After pilot study
• Sample size may be reduced
• Method of collection of information may be
  changed
• Sample design may be changed
• Questionnaire is revised
• Questionnaire designers must understand the need
  to pretest, pretest and pretest ASA

23
Non-Response those who do not reply
4. Non-Response
What can go wrong?

• Refusals
• Not available at time of call
• Questionnaire too difficult

Non-respondents usually differ from respondents
Low response will make sample unrepresentative
24
Reducing non-response
5. Non-Response
What can go wrong?

• Short questionnaire
• Sponsorship by bodies which are highly regarded
• Assuring confidentiality
• Start the questionnaire with non-threatening
  questions
• Rewards
• Offer to provide final results later on
• Call backs

25
Reducing Non-Response Call backs/Follow ups
26
Sample Size
5. Sample size too large
What can go wrong?

• Sample size is determined by
• Cost constraints
• Time constraints
• Resources available
• The size of the population is usually not
  important when deciding the sample size.
• How the population varies is more important.

27
Frog PopulationsEstimate sample size
necessary to calculate average weight.
Size of Sample?
1
Sample Size
28
Frog PopulationsEstimate size necessary to
calculate average weight.
Size of Sample?
gt1
Sample size
29
What every survey should have

• Clearly defined objectives
• An adequate frame
• Pilot study
• Follow ups
• Response rate at least 70
• Small sample

30
Analysis

• Excel
• SPSS
• Library packages , eg. Libra, Websurveyer

31
Recommending reading for those considering doing
a survey

• What is a survey?
• How to plan a survey
• How to collect survey data
• How to conduct pretesting
• Mail surveys
• Telephone surveys
• Designing a questionnaire

www.amstat.org/sections/srms/brochures/ Survey
Research Methods, American Statistical
Association