- Foundation
- About 20 min
- 5 practice questions
Sampling errors, non-sampling errors and agencies that collect data
Calculate a sampling error, explain the kinds of non-sampling error and why they are more serious, and describe the Census of India and the National Sample Survey.
If something looks wrong, please tell us.
By the end of this lesson you can
- Define and calculate sampling error.
- Explain sampling bias, non-response and errors in data acquisition.
- Describe the work of the Census of India, the NSS and other national agencies.
A sample is only a part of the population, so its results will usually differ a little from the truth. Some errors come from sampling itself; others come from the way data are gathered and recorded. This lesson separates the two and ends with the agencies that collect India's official data.
Population parameters and sample estimates
A population of numerical values has two important characteristics: central tendency (measured by the mean, median or mode) and dispersion (measured by the range, mean deviation, standard deviation and so on). The true value of such a characteristic for the population is a parameter; the value worked out from a sample is an estimate.
Sampling error is the difference between the actual value of a population parameter and its estimate from a sample. It can be reduced by taking a larger sample.
Worked example 1: calculating a sampling error
Illustrative: a made-up population of six weavers has daily earnings (₹) of 320, 360, 400, 440, 480 and 400. Step 1, population mean (parameter): 320 + 360 + 400 + 440 + 480 + 400 = 2,400; 2,400 ÷ 6 = ₹400. Step 2, a sample of two (320 and 360): mean = (320 + 360) ÷ 2 = 680 ÷ 2 = ₹340. Sampling error = 400 − 340 = ₹60.
Step 3, a larger sample of four (320, 360, 400, 480): mean = 1,560 ÷ 4 = ₹390. Sampling error = 400 − 390 = ₹10. Check: 320 + 360 + 400 + 480 = 1,560. The larger sample came much closer to the parameter, which illustrates why sampling error can be reduced by taking a larger sample (though any single sample may still be off).
| Sample (illustrative) | Size | Sample mean (₹) | Sampling error (₹) |
|---|---|---|---|
| 320, 360 | 2 | 340 | 60 |
| 320, 360, 400, 480 | 4 | 390 | 10 |
| All six (census) | 6 | 400 | 0 |
Non-sampling errors
Non-sampling errors are more serious than sampling errors, because they cannot be reduced simply by taking a larger sample. Even a census can contain them. The chapter describes three kinds:
Kinds of non-sampling error
-
Sampling bias
Some members of the target population could not possibly be included
- Example: a phone survey that leaves out people without phones
-
Non-response error
A listed person cannot be contacted or refuses to answer
- The remaining sample may not be representative
-
Errors in data acquisition
Incorrect responses are recorded
- Different measuring tapes give different lengths
- Prices differ by shop and quality, so only averages can be used
- Recording or copying mistakes, such as 31 written as 13
Worked example 2: the effect of a recording mistake
Illustrative: the prices (₹ per kg) of oranges in five made-up shops are 60, 64, 70, 66 and 80. True average: 60 + 64 + 70 + 66 + 80 = 340; 340 ÷ 5 = ₹68. Suppose the enumerator writes 08 instead of 80. Recorded average: 60 + 64 + 70 + 66 + 8 = 268; 268 ÷ 5 = ₹53.60. Error caused: 68 − 53.60 = ₹14.40. Taking more shops would not cure a mistake like this; only careful recording and checking can, which is why non-sampling errors are called more serious.
Sampling and non-sampling errors
-
Sampling error
- Arises because only a part of the population is studied
- Difference between parameter and estimate
- Falls as sample size rises
- Absent in a census
-
Non-sampling error
- Arises from bias, non-response or faulty data acquisition
- Not cured by a larger sample
- Can occur even in a census
- More serious
Agencies that collect data in India
Several national and state agencies collect, process and tabulate statistical data. The textbook names, at the national level, the Census of India, the National Sample Survey (NSS), the Central Statistics Office (CSO), the Registrar General of India (RGI), the Directorate General of Commercial Intelligence and Statistics (DGCIS) and the Labour Bureau. (Names and structures of agencies can change over time; these are the ones listed in the book.)
| Agency | What the textbook says it does |
|---|---|
| Census of India | Gives the most complete and continuous demographic record of the population; described as conducted every ten years since 1881, with the first after Independence in 1951; covers size, density, sex ratio, literacy, migration and rural-urban distribution |
| National Sample Survey (NSS) | Set up by the Government of India for nation-wide surveys on socio-economic issues, in successive rounds; reports and the journal Sarvekshana; estimates of literacy, school enrolment, employment, unemployment, enterprises, health, use of the public distribution system; also industrial activity and retail prices, used for planning |
Data published by these agencies are secondary data for the students and researchers who use them.
Key takeaways: sampling error is the gap between a population parameter and its sample estimate and shrinks as the sample grows. Non-sampling errors (sampling bias, non-response, errors in data acquisition) are more serious because a bigger sample does not remove them. The Census of India and the NSS are major national sources of data on population and socio-economic conditions.
Check your understanding
-
1. Illustrative: the true average income of a population is ₹750, and a sample gives an average of ₹712. The sampling error is:
- ₹38
- ₹712
- ₹1,462
- ₹62
Show answer
Answer: A. Sampling error = parameter − estimate = 750 − 712 = ₹38.
-
2. Which error can be reduced by taking a larger sample?
- Non-response error
- Sampling error
- Recording error
- Sampling bias
Show answer
Answer: B. Sampling error arises from studying only part of the population and falls as the sample becomes larger. The other three are non-sampling errors, which a larger sample does not cure.
-
3. An enumerator records a household's size as 13 when the true figure is 31. This is an example of:
- sampling error
- sampling bias
- error in data acquisition
- non-response
Show answer
Answer: C. Recording or copying the wrong figure is a mistake in data acquisition, a kind of non-sampling error.
-
4. According to the textbook, the journal through which NSS data are released is:
- the Census Gazette
- the Labour Bulletin
- the Statistical Abstract
- Sarvekshana
Show answer
Answer: D. The chapter states that NSS data are released through reports and its quarterly journal Sarvekshana.
-
5. Assertion (A): Non-sampling errors are more serious than sampling errors. Reason (R): Non-sampling errors can occur even in a census and cannot be removed simply by increasing the sample size.
- Both A and R are true, and R is the correct explanation of A
- Both A and R are true, but R is not the correct explanation of A
- A is true, but R is false
- A is false, but R is true
Show answer
Answer: A. Both are true, and R explains A: since a larger sample or even a full census does not remove them, non-sampling errors are harder to control and so more serious.
Sign in to keep track of the lessons you finish.