Non-Sampling ErrorNon-Sampling Error
Sampling and Non-Sampling ErrorSampling and Non-Sampling Error
 Sampling error arises due to the sampling
technique used to draw a sample.
Non-Sampling error is due to the reasons
other than sampling techniques during a
survey process.
Both the errors can be controlled but
cannot be made zero.
Complete enumeration is free from sampling
errors but is subject to non-sampling error.
Sample survey is subject to both sampling
and non-sampling error.
Non-Sampling ErrorNon-Sampling Error
 Sampling error decreases with increase in sample size, whereas non-
sampling error increases with increase in sample size.
 Non-sampling error is likely to be more in case of complete enumeration
survey than in case of a sample survey.
 Non-sampling error can occur at every stage of planning and execution of
a census or a survey. Such as:-
 Faulty definition
 Improper Training methodology
 Defective methods of data collection
 Incomplete coverage or non-response
 Improper or ambiguous questionnaire
 Lack of monitoring or supervision of primary staff
 Errors during tabulation
 Errors in data processing operations
 Improper interpretation during analysis and report writing
Non-Sampling ErrorNon-Sampling Error
Contd…Contd…
Major source of non-sampling error can
be divided into two groups.
1. Non-observation error
2. Measurement or observation error
Non-Observation ErrorNon-Observation Error
Non-Observation Error
Two sources of Non-observation error
are most important
1. Non-coverage
2. Non-response
Non-Observation ErrorNon-Observation Error
Contd…Contd…
 Non-coverage occurs when there are units in the population
of interest that have no chance of being sampled for survey.
 Possible reasons:
 Geographical limits like hilly areas, boundaries, etc.
 Institutional restrictions like military establishments, jails, etc.
 Incomplete frame differing from target population.
 Deliberate leaving of a portion of population out of frame,
Ex-A restriction to travel to a particular region due to riot
or civil unrest.
 Cost factor
 Use of out of date or inaccurate household listing as frame.
Non-Observation ErrorNon-Observation Error
Contd…Contd…
 Non-coverage arises due to errors or problems in
the frame being used for sample selection; but non-
response arises after frames have been constructed
and sample elements selected from the frame.
 Possible measures to control non-coverage error:
Reducing the level of non coverage through improved
field procedure like use of multiple frames or through
improving the listing process.
Clearly defining the target population.
Use of up to date more accurate household listing as
frame.
Non-Observation ErrorNon-Observation ErrorContd…Contd…
 Non-response occurs when a sampled unit fails to participate
in the survey, either completely or partially. Non-response is
believed to be more in household surveys.
 Non-response can be divided in two categories.
1. Unit non-response
2. Item non-response
 Unit Non-response
 Unit non-response can be at primary or secondary stage
sample units.
 Due to weather condition or natural disaster an entire 1st
stage unit may become non-response.
 A listed household may become non-responsive due to non-
cooperation of household member or the household
member may not be available.
Non-Observation ErrorNon-Observation Error
Contd…Contd…
 Unit Non-response
 Non-response of household unit may be due to following
reasons.
◦ Refusal to participation.
◦ Non-availability during survey period.
◦ Language barrier.
◦ Health condition, etc.
 Remedy
◦ Household member may be persuaded for giving reply.
◦ Frequent visit
◦ Awareness campaign for improving rate of participation.
◦ Increasing sample size to adjust non-response base upon the
past experience.
Non-Observation ErrorNon-Observation Error
Contd…Contd…
 Item Non-response
 Item response refer to the situation where some items of the
questionnaire are not answered by the respondent.
◦ Female health related questions.
◦ Questions related to disability.
◦ Questions related to income.
◦ Questions related to property, etc.
 Remedy
◦ Predicting missing values based on related trend analysis or from
other available information.
◦ Interviewers can be trained to get answer by cross questioning or
seeking for other related information from which the required
answer can be predicted.
◦ Interviewer may try to collect information in a range.
Measurement ErrorMeasurement Error
 Measurement error occurs when there is a difference
between the respondent’s answer to a question and
the true value.
 The measurement error is otherwise known as
observation error.
 Sources of Measurement error
 Effect of questionnaire design
 the topic it covers
 wording of the questions
 Length of the questions
 Length of the questionnaire
 Order of the questions
Measurement ErrorMeasurement Error
Contd….Contd….
Sources of Measurement error
 Mode of data collection
 Face to face interview: Use electronic mode like Tabs/
Palmtops improves quality of data collection and
reduces error than paper and pencil.
 Presence of outside members may affect the answers
 Direct Observation
 Interviewer effect
 Interviewer error can be reduced through interviewer
training, supervision and workload manipulation.
 Respondent effect
Measurement ErrorMeasurement Error
Contd….Contd….
Effective Training can reduce Non-
sampling error
 There should be less layers of training
 Uniform module should be developed so
that every trainer and every trainee should
talk the same language.
 Manual or module should have empirical
examples or filled in schedules.
 FAQ and common mistakes committed
should be part of training manual.
Measurement ErrorMeasurement Error
Contd….Contd….
Effective monitoring mechanism can
reduce Non-sampling error
 Sample selection for supervision should be
done at central level.
 Higher layer of supervision should be done
from among the units already checked at
lower level.
 Supervision should be minimum two levels.
 Technical supervision should be done on
some key indicators.
THANK YOU

Non sampling error

  • 1.
  • 2.
    Sampling and Non-SamplingErrorSampling and Non-Sampling Error  Sampling error arises due to the sampling technique used to draw a sample. Non-Sampling error is due to the reasons other than sampling techniques during a survey process. Both the errors can be controlled but cannot be made zero. Complete enumeration is free from sampling errors but is subject to non-sampling error. Sample survey is subject to both sampling and non-sampling error.
  • 3.
    Non-Sampling ErrorNon-Sampling Error Sampling error decreases with increase in sample size, whereas non- sampling error increases with increase in sample size.  Non-sampling error is likely to be more in case of complete enumeration survey than in case of a sample survey.  Non-sampling error can occur at every stage of planning and execution of a census or a survey. Such as:-  Faulty definition  Improper Training methodology  Defective methods of data collection  Incomplete coverage or non-response  Improper or ambiguous questionnaire  Lack of monitoring or supervision of primary staff  Errors during tabulation  Errors in data processing operations  Improper interpretation during analysis and report writing
  • 4.
    Non-Sampling ErrorNon-Sampling Error Contd…Contd… Majorsource of non-sampling error can be divided into two groups. 1. Non-observation error 2. Measurement or observation error
  • 5.
    Non-Observation ErrorNon-Observation Error Non-ObservationError Two sources of Non-observation error are most important 1. Non-coverage 2. Non-response
  • 6.
    Non-Observation ErrorNon-Observation Error Contd…Contd… Non-coverage occurs when there are units in the population of interest that have no chance of being sampled for survey.  Possible reasons:  Geographical limits like hilly areas, boundaries, etc.  Institutional restrictions like military establishments, jails, etc.  Incomplete frame differing from target population.  Deliberate leaving of a portion of population out of frame, Ex-A restriction to travel to a particular region due to riot or civil unrest.  Cost factor  Use of out of date or inaccurate household listing as frame.
  • 7.
    Non-Observation ErrorNon-Observation Error Contd…Contd… Non-coverage arises due to errors or problems in the frame being used for sample selection; but non- response arises after frames have been constructed and sample elements selected from the frame.  Possible measures to control non-coverage error: Reducing the level of non coverage through improved field procedure like use of multiple frames or through improving the listing process. Clearly defining the target population. Use of up to date more accurate household listing as frame.
  • 8.
    Non-Observation ErrorNon-Observation ErrorContd…Contd… Non-response occurs when a sampled unit fails to participate in the survey, either completely or partially. Non-response is believed to be more in household surveys.  Non-response can be divided in two categories. 1. Unit non-response 2. Item non-response  Unit Non-response  Unit non-response can be at primary or secondary stage sample units.  Due to weather condition or natural disaster an entire 1st stage unit may become non-response.  A listed household may become non-responsive due to non- cooperation of household member or the household member may not be available.
  • 9.
    Non-Observation ErrorNon-Observation Error Contd…Contd… Unit Non-response  Non-response of household unit may be due to following reasons. ◦ Refusal to participation. ◦ Non-availability during survey period. ◦ Language barrier. ◦ Health condition, etc.  Remedy ◦ Household member may be persuaded for giving reply. ◦ Frequent visit ◦ Awareness campaign for improving rate of participation. ◦ Increasing sample size to adjust non-response base upon the past experience.
  • 10.
    Non-Observation ErrorNon-Observation Error Contd…Contd… Item Non-response  Item response refer to the situation where some items of the questionnaire are not answered by the respondent. ◦ Female health related questions. ◦ Questions related to disability. ◦ Questions related to income. ◦ Questions related to property, etc.  Remedy ◦ Predicting missing values based on related trend analysis or from other available information. ◦ Interviewers can be trained to get answer by cross questioning or seeking for other related information from which the required answer can be predicted. ◦ Interviewer may try to collect information in a range.
  • 11.
    Measurement ErrorMeasurement Error Measurement error occurs when there is a difference between the respondent’s answer to a question and the true value.  The measurement error is otherwise known as observation error.  Sources of Measurement error  Effect of questionnaire design  the topic it covers  wording of the questions  Length of the questions  Length of the questionnaire  Order of the questions
  • 12.
    Measurement ErrorMeasurement Error Contd….Contd…. Sourcesof Measurement error  Mode of data collection  Face to face interview: Use electronic mode like Tabs/ Palmtops improves quality of data collection and reduces error than paper and pencil.  Presence of outside members may affect the answers  Direct Observation  Interviewer effect  Interviewer error can be reduced through interviewer training, supervision and workload manipulation.  Respondent effect
  • 13.
    Measurement ErrorMeasurement Error Contd….Contd…. EffectiveTraining can reduce Non- sampling error  There should be less layers of training  Uniform module should be developed so that every trainer and every trainee should talk the same language.  Manual or module should have empirical examples or filled in schedules.  FAQ and common mistakes committed should be part of training manual.
  • 14.
    Measurement ErrorMeasurement Error Contd….Contd…. Effectivemonitoring mechanism can reduce Non-sampling error  Sample selection for supervision should be done at central level.  Higher layer of supervision should be done from among the units already checked at lower level.  Supervision should be minimum two levels.  Technical supervision should be done on some key indicators.
  • 15.