Sampling refers to the act of suitably selecting some units from a population under study. The sample is chosen so that it is a good representative of the population. Then the characteristics of the population can be studied by estimating their values on the basis of the sample. We now list out some of the merits and limitations of using sampling methods in statistics.

**Advantages of Sampling:**

There are many advantages of sampling over the 100% inspection method. Some of them are:

- The sampling method is much more faster than the complete ennumeration method. We save time because not only because there is much less data to collect but also because much less time is needed to analyse and interpret the data. Thus, sampling methods are of great use in cases where data is urgently required.
- Sampling is significantly cheaper than the method of 100% inspection beacuse there are lesser number of units to study. In cases where there may be a lack of funds, sampling allows us to generate conclusions accurately and cheaply.
- Since the sampling survey is conducted under the supervision of experts the conclusions generated are reliable. The expert supervision ensures that there are no non-sampling errors. The sampling errors can also be reduced by suitable choice of sampling method.
- Setting up a complete ennumeration census requires a lot of organizational effort at a large scale. A large number of ennumerators must be hired and they must co-ordinate with each other. On the other hand, sampling studies are easier to set up and organize since fewer people are involved.
- If there are non-responses or incomplete responses, then it is much easier to follow up on those few individuals in the sample and get the correct response. In a complete census, there may be too many non-response or incomplete responses than manageable.
- Sampling allows a deeper study of some aspects of the data. Suppose you have a fixed amount of time and money. Rather than asking 15 questions to every individual, it is much better to ask 50 questions to a representative sample of them. Therefore, sampling allows us to make a deeper study of the characteristics of the population under given constraints of time and money.
- If the population is infinite or very large then sampling is the only way to study such populations. Consider the example of throwing a dice. Theoretically a dice can be thrown infinite number of times. But if we want to decide whether the coin is unbiased or not it is enough to throw the dice some a finite large number of times. If we get roughly equal number of heads and tails in these finite number of throws we conclude that the dice is unbiased. The conclusion that the dice is unbiased is made on the basis of the sample of finite throws.
- Suppose you want to test the stregth of chalk produced by a particular company. Here the testing involved is destructive since it damages the unit under study. In such circumstances we cannot possibly test each chalk. Therefore in cases of destructive testing, sampling is the only method of study available.

**Disadvantages of Sampling:**

- If the researcher is biased and does not choose a random sample then the sample will not be a good representative of the population.
- Since sampling studies requires experts who are well trained, it is much more difficult to find competent manpower.
- Even if the study is well-designed, there is no way to completely eliminate the chances of sampling errors. There is always a chance (although small) that the sample chosen is not representative.
- Sometimes we want to know information about each and every unit in the population. For example, when a country conducts population census, each and every person is counted. Sampling methods are not helpful in such situations.
- If the sampling procedure is not properly planned and executed then the conclusion become unreliable.
- One of the limitations of sampling in marketing research is that when conducting marketing research some people in the sample may refuse to respond. Removing or replacing such units intoduces error into the study.