Sampling Theory

Only available on StudyMode
  • Download(s) : 179
  • Published : May 22, 2013
Open Document
Text Preview
Session 5 Topic: Sampling Theory/ Techniques and Discussions



Project Brief
◦ Expectations and deliverables
 (Deadline October 1, 2010- EOD)



Sampling basics
◦ Fundamental Issues ◦ Errors



Sampling techniques
◦ Probabilistic ◦ Non-probabilistic



Discussions

© Krishanu Rakshit, IIM Calcutta

28 September, 2010

2

  

When do we use a ‘sample’ When do we use a census (population) Sampling errors ◦ Difference between a measure from sample and the measure which can be obtained from the population



Non-sampling errors
◦ Selection Error ◦ Population specification Error
 A bias/error which creeps in when sample obtained through nonprobabilistic techniques does not represent the population  When an inappropriate population is chosen from which a sample is selected  E.g. could be choosing a sample of cat owners for researching ‘dog food ‘

© Krishanu Rakshit, IIM Calcutta

28 September, 2010

3



Non-sampling errors
◦ Sampling frame error
 A sampling frame is a directory (defining the population from which sample will be drawn)  The error comes in when the sample is drawn from an inaccurate sampling frame

◦ Surrogate Information error

 E.g. for studying the preference for desktops at home, the sampling frame is considered to be ‘subscription list of PC World magazine’

 Often information sought is different from the information required- one of the critical errors  E.g. Classic Case of new Coke- should not have been consumers’ preferences between Old and new Coke, rather a measure of attitudinal aspects of a change in taste

© Krishanu Rakshit, IIM Calcutta

28 September, 2010

4



Non-sampling error
◦ Measurement error
 Difference between information sought and information collected by the procedure (problem with Questionnaire design)  Validity issues

 Experimental error
 Mainly from faulty experimental design  Critically, from interaction effects due to experiments (CG, EG checks)

 Data Analysis Error
 Error in collecting, coding, analysing and interpreting of Data  Usually minor

© Krishanu Rakshit, IIM Calcutta

28 September, 2010

5



Non-sampling errors
◦ Response Errors
 Or, data errors, when respondents provides inaccurate answersoften due to lack of comprehension

◦ Non-response errors
 Due to non-selection- some members may not be contacted; so responses not included  Others providing incomplete answers for the questionnaire

◦ Other minor non-sampling errors would be:
 Administering errors- Questioning error, recording error, interference (interviewer fills in)

© Krishanu Rakshit, IIM Calcutta

28 September, 2010

6

Total Error

Sampling Error

Non Sampling error

Design Errors

Administering Errors

Response Errors

Non-response Errors

Selection, Population Specification, Sampling Frame, Surrogate Information, Measurement Error, Experimental Error

Questioning error, Recording Error, Interference Error

Data Error, Which could be intentional, Unintentional

Failure to contact all, Incomplete responses

© Krishanu Rakshit, IIM Calcutta

28 September, 2010

7



It is critical to determine the target population

◦ To eliminate the Specification error as well as sampling frame error ◦ A cliché, however, definition of Research objective is critical ◦ Optimal definition of population  Most research failures suffer from this problem!  This clarity (or lack of it) impacts the questionnaire design  Too fine and it is too restrictive, expensive and operationally difficult  Too broad and might confound the findings  But convenience not at the expense of appropriate research design

◦ Convenience is critical

© Krishanu Rakshit, IIM Calcutta

28 September, 2010

8



Sampling frame
◦ Selection of the list
 Telephone Directory
 In most cases, it works fine as it provides a complete list.  Sometimes it may not be...
tracking img