METHODS OF DATA GENERATION
SAMPLE SURVEYS
SAMPLE SURVEYS Domesday book was a famous census of England conducted in 1085 to 1086 by William the conqueror
SURVEY= an investigation of the behaviour/opinions of th Group of people or a sample
Sample=one part of a whole that can be examined in ord Know what the rest is like Eg., technology survey on marginal farmers
Research questions appropriate for a survey?
•Behavior •Attitude/beliefs/opinions •Characteristics •Expectations •Self classification •knowledge
Experiment: 2. subjects respond to treatment created by the researcher 3. Causality is shown by timing of treatments 4. Associations between treatments are observed 5. Controlling alternative explanations 6. Physical control on treatments
Survey: 2. Sample many respondents and ask the same question 3. Measure many variables 4. Test many hypothesis 5. Infer temporal order from questions about past behavior 6. Survey researchers measure variables which represent alternative explanations 7. Survey research is correlational 8. Approximations of the rigorous test are done
Types of surveys
1. Mail and self administered questionnaires 2. Telephone,email,computer based interviews/surveys 3. Face to face interviews 4. Panel technique-select group and study attitude(opinion poll) 5. PRA and special situations
Steps in sample surveys
4. Design and planning phase 5. Data collection phase 6. Data analysis and presentation
1. Design and planning phase • Objectives of the study
• Decide on type of survey, respondent and pop
1. Design and planning phase
•Develop the survey instrument/questionnaire/semi Structured interview schedule(PRA Write questions to measure variables Decide on response categories Organise question sequence Design questionnaire layout
1. Design and planning phase
•Plan a system for recording answers,items of informn •Pilot test the instrument and train the interviewers
1. Design and planning phase (sampling) • Draw the sample by design or sampling design Define target population Decide on type of sample Develop sampling frame Decide on sample size Select a sample
Sampling
Universe
draw
Sample which represents the population
Why should we sample for a survey?ROLE • make sample more efficient •Manageable •Cost effective-Provide for lowest possible Cost
Terms used in sampling surveys
Population: a large pool of cases or elements(all the cases/elements)/even the term universe is used Target population: specific pool of cases with a criteria Sampling frame ; a specific list of all cases matching the criteria Sampling element: unit of analysis or case Sampling ratio: is the ratio of the sample to the
Terms used in sampling surveys Sampling design= Rule to draw the sample Random sample= is not a haphazard sample but based On the laws of probability
How large should your sample size be? •Smaller the popn bigger the sample size and vice versa •1000 popn-30 percent should be the sample •10,000 popn 10 percent should be the sample •For large population 1,50,000 –1 percent sample •For very large population(10 million and
Principle of sampling
Small sample –small increase produce big gains 50-100-errors reduce to 2.1% from 7.1% 1000-2000-errors reduce from 1.6-1.1%
Sample size depends upon
1. Degree of accuracy 2. Degree of variability 3. Number of variables examined
When above increases increase sample size
NON PROBABILITY SAMPLE
(no sampling frame)
PROBABILITY SAMPLE
(sampling frame)
Haphazard: select anyone convenient
Simple: true random procedure
Quota: select anyone in predetermined groups
Systematic ; select every kth person
Purposive ; select anyone in hard to find target popn
Stratified ; randomly select people in predetermined groups
Snowball: select people connected to one another
Cluster : take multistage random sample in several levels
NON PROBABILITY SAMPLE
Haphazard sample : • select anyone convenient •Seriously misrepresents the popn
When to use: For entertainment value(MTV Bakra)
Shortfall: Distorted view Seriously misrepresent the popn
NON PROBABILITY SAMPLE
Quota sample : • identify categories (male/female) •5 males 5 females under 30 years of age •Then use haphazard sampling When to use: For entertainment value(MTV Bakra)
Shortfall: Distorted view Seriously misrepresent the popn
NON PROBABILITY SAMPLE
Purposive sample : • judgemental •Uses judgement of a n expert •Informative cases When to use: When difficult to reach To identify a case in depth
Shortfall: Cannot generalize for a population Purely opinionistic
NON PROBABILITY SAMPLE
Snowball sample : • interconnected networks •sociometry •Key informants When to use: Used in PRA and RRA
Shortfall: Cannot generalize for a population Purely opinionistic Has temporal and spatial limitations
PROBABILITY SAMPLE
Random sample : • use random number/table •With replacement •Without replacement When to use: When population is well defined When sampling frame is established
Shortfall: Cannot generalize for various groups
PROBABILITY SAMPLE
Systematic sample : • number the sample •Select the kth elements
When to use: When population is well defined When sampling frame is established
Shortfall: Cannot generalize for various groups
PROBABILITY SAMPLE
Stratified sample : • divide the popns into subpopns •Subpopns=strata •Random sample from sub popn using random or systematic samplng When When When When Shortfall: Time consuming costly
to use: variation is there in the strata population is well defined sampling frame is established
PROBABILITY SAMPLE
Cluster sample : • when the population is dispersed spatially on a large area •Randomly select districts then taluks then villages(3 stage)then farmers •When clusters list is available When to use: When population list is not there But the list of clusters(subgroups are there) When sampling frame is established Shortfall: Time consuming Requires lot of information
PROBABILITY SAMPLE
Random digit dialing sample : • you can stratify/cluster/random sample also •Uses the telephone directory •Population is the telephone directory When to use: used for telephone interviews When sampling frame is established
Shortfall: Not so much used in India and in agricultural situations May suit only big farmers
Why sample survey predictions are wrong 1936-F.D Roosevelt won by a landslide but Alf landon was Predicted-mistakes in sampling(sampling frame was wrong)
It excluded people without telephones or automobiles lower Income favoured roosevelt 1948-gallup predicted the wrong candidate
1. Data collection phase • Organize the survey • Locate and contact the respondents
• Make introductory statements or provide instructio • Ask questions and record answers
• Thank respondent and continue to next responden • End data collection and organize data
1. Data analysis and presentation • Tabulate properly using MS excel,any spreadsheet • Use appropriate statistical tools • Proper interpretation using your subject and experience
Advantages
1. Great deal of information can be collected 2. Large population can be used 3. Reduces labour 4. Information is accurate 5. Variety of information can be collected
Disadvantages
1. In-depth study not possible 2. Costly 3. Time consuming 4. Scope for sampling error