13.1 MEANING AND OBJECTS OF ‘SAMPLING’ ng’ denotes the selection of a part of the aggregate Statistical view to obtain information about the whole. This aggregate or totali information on a particular character of all the members covered bya is called Population or Universe. The selected part, which is used 0 ascertain th characteristics of population is called sample (Section 1.4, Page 6). While taking” sample, the population is assumed to be composed of individual units or mene, some of which are included in the sample. The total number of Members of the population and the number included in the sample, called ‘population size’ and “sample size’ respectively. A statistical population is ‘finite’ or ‘infinite’ according to its size. When the number of members of the population can be expressed as a definite quantity, the population is said to be ‘finite’. Otherwise the population is ‘infinite’ , In particular, if a sample is known to have been drawn from a continuous probability distribution, then the population is infinite. A complete list of all the units in finite Population, properly numbers for identification, is called a Sampling Frame, €.g., a list of all households in @ given region as used in population census. Again, the population may be ‘existent’ or ‘hypothetical’ . The population of incomes of all persons in a country is an example of existent population; because the member of the population really exist, though they may not be actually known because o! practical difficulties. The population of points obtained in all possible throws of ade is an example of ‘hypothetical’ population. , Ifa sample of an adequate size is properly chosen and analysed, it is most likely : reveal the characteristics of the whole population, and the results obtained thee can be fairly relied upon as if they were based on all members of the population of? possibility of reaching valid conclusions concerning a population by mat Properly chosen sample is based on two important laws: (i) Law of as Regularity, and (ii) Law of Inertia of Large Numbers. The former states that te of reasonably large size when selected at random, is almost sure t0 ee sl characteristics of the Population (page, 11). The second law states that a cel large size show a high degree of stability, i.e., the results obtained theor’™ to be very close to the population characteristics. Materia With LY Of statistic | in NVESti pationRe vantages of sample survey over complete enumeration A survey conducted on a Suitable samy enumeration preferred ‘to ple is called a sample survey, Of complete census the entire populaton is surveyed. General: complete enumeration for the following reasons : {a) Reduction of cost : hie in comply e ele * SAMPLE survey The cost, either in terms than ina complete census. the total cost is likely of money or in terms of man-hours, is less in a Sample surjey Although the cost per unit is usually greater in a Sample surye, to be smaller because a sample comprises only a Part of the Population, (b) Greater speed : Since a sample comprises only some of the population members, and processed more rapidly in sampling than in complete enw consideration when the information are urgently needed. » the data can be collect meration. This is a vi (c) Greater scope as the smaller scale of a sampi snumerators enable the enquirer to collect more information and to include More areas in the survey. (d) Greater accuracy + A sample survey generally gives more accurate and teliable data than a complete census due to the possibility of employing better-trained Personnel, effecting better Supervision or using better equipments. ‘pected to furnish correct information Teadily when they know that they have been selected from the whole population for the Study. The combined effect of sampling and non-sampling errors in sample survey is likel . : : ¥ t© be much less than the effect of non-sampling errors in complete enumeration, The error arising due to drawing inferences about the population on the basis of sample is termed sampling error, and the errors ma inly arising at the Stages of collection and processing of data are termed non-sampling errors, Non-sampling en are common i b cn complete enumeration and sample surveys. Generally, the Non-sampling error increase ioe ae ling error decreases with increase in sample size. Feases, but the sampling (e) Measure of accuracy : It should also be noted that complete census general] sampling errors and there is no way of meas Y involves a large f non eg the maon:, ge amount o! j of m ing the Magnitude of Such errors, On the oll! hand, a properly selected sample will give an idea of the magi ; involved in the estimates. Nitude of the sampling error13.5 Biases in surveys Bias is an effect that deprives a statistical result of representativeness by systematically distorting it. It is different from a tandom error which may distort the result on any occasion put balances out on the average. The different types of bias that are likely to arise in a survey may be broadly classified under two heads : . (1) Procedural biases, and 2) Sampling biases. (1) Procedural Biases : These arise in a complete census as well as in a sample survey. The following are the various kinds of such biases : (i) Response biases : These biases arise from wrong responses of the informants. For instance, one may deliberately Over-state one’s education or under-state one’s age. In many cases, one may furnish wrong answers to guard one’s self-interest; such as, a person may give sm under-statement of his income, personal savings, etc. This type of bias may also arise due ‘0 preference of the people for certain numbers, like multiples of 5 or even numbers. (ii) Interviewer biases : Answers given by the informants after receiving suggestions from ‘te interviewer are likely to be affected by the interviewer's own beliefs and prejudices. (ill) Non-response biases : These may arise in a survey if the respondent is not found at ‘ome even after repeated calls or if he either fails or refuses to answer certain questions. liv) Observational biases : In case of finding the variable value by observation, biases may ‘Ccur due (0 the influence of psychological factors. For instance, people with different political ‘iliations may say widely different figures for sive of the gathering in an open meeting of a tltical panty, 8 Sa pling biases : These. occur only in sampling. Here the following types may be nised, i! Biases due to defective sampling method : If a proper random procedure is not strictly bee » the investigator may use his own judgement or discretion in selecting the units, and, % introduce bias. niinkatnssssnnninsainiananitit assassin(id) Biases due to substiution : When difficulties arise in enumerating a member o: included in the sample, investigators generally substitute it by another convenient lly Clearly, this leads to some bias due to the difference in the characteristics of the yy and the original members. stitute (ii) Biases due to faulty demarcation of sampling units : In a crop-cutting sury example, there is an inclination on the part of the investigators to include some ie for | im the sample when they are situated near but outside the selected area; this results inp estimation. Over. (iv) Biases due to wrong selection of the statistic : For estimating a parameter, one sho ld use the appropriate statistic; otherwise, a constant bias may occur. For example, sample vai ann is a biased estimator of population variance in simple random sampling with replacement, yes of population and types of sampling A population may be finite or infinite depending on whether it contains a finite or an infinite number of members. For example, the population of houses in a certain locality is finite, while the population of atmospheric pressures at different points in a room is infinite. Again, the population may be existent or hypothetical. If the population consists of concrete existent individuals, it is called an existent population. On the other hand, if the individuals of the population do not exist in reality, but exist only in imagination, then the population is inypothetical. For example, the outcome of an infinite number of tosses of a coin represents a hypothetical population of heads and tails. Sampling is broadly classified as subjective and objective. In subjective sampling, the selection of members depends on the personal judgement or discretion of the sampler, while in objective sampling, there is a specific rule of selection and the selection is independent of the judgement or discretion of the sampler. It should be noted that any haphazard or motivated selection will lead to subjective sampling and such a sampling generally involves a bias of unknown magnitude. Objective sampling, in its turn, is classified as non-probabilistic, probabilistic (ot random) and mixed. When there is no probability assigned to the mode of selection but the sampling done according to a specific rule, then the sampling is called non-probabilistic. In random sampling, each member has a definite preassigned probability of being selected. Again. if the sampling is partly probabilistic and partly non-probabilistic, it is called mixed samplins, Suppose, for instance, that we are to select 10 trees from a row of 80 trees in fores Starting with the first tsee, if we select every 8th tree, then sampling becomes non-probabilis® If 10 trees are so selected that each of 80 trees has a definite preassigned probability of bei ees the sale then the sampling is probabilistic, If, however, one tree is oO Fan from the first 8 trees, and then every 8th tree i i the select’? ° the samigling becomes ‘nixed. ry (ree is taken starting from