Stratified Sampling

In stratified random sampling the population is divided in to sub-population i.e. two or more strata and then the units are drawn form each stratum. A simple random sample is taken form each stratum and the sub samples are joined together rot constitute the total sample. The universe can be divided on the basis of sex-male, female; age-young, middle aged, and  old; income, ovulation, rural-Urban back ground and so on. For example if we want to study the family structure in a particular area, the universe can be stratified on the basis of income (upper, middle and low income group) and rural-urban back ground. This would yield six strata, and the unties are selected at random from each stratum. The process of selection in each stratum is carried out separately and independently.

Stratified sampling may be proportional or disproportional. The customary and frequent practice is to draw proportionate stratified sampling; the units from the stratum are drawn in the same proportion as income group is 20%in the universe and the urban lower income group is 30%, the same proportion will be selected in the sample also. In disproportional stratified sampling, varying proportion of units are drawn form all the strata, irrespective of their percentage in the Universe.

According to Jahoda, sampling the various strata in different proportions may be due to several reasons. Sometimes it is essential to increase the proportion of strata having small numbers, in order to ensure that these classes are sampled at all and to enable meaningful comparison, strata one stratum is much more homogeneous with respect to some characteristic being studied than the other. Jahoda writes ‘For a given degree of precision, it will take a smaller number of cases to determine the state of affairs in the fist stratum indicate that with respect certain types of attitudes and opinion, men will different much more than women One would accordingly plan the sample in such a way so as to include more men among the men. Since women may expected to be homogeneous. They do not have to be sampled as thoroughly as men for a given degree of precision.

Another major reason for varying the-sampling proportion is that large number of cases is selected form one stratum then from another because one many want to subdivide the cases within one stratum for further analysis. If a given stratum has relatively few cases, sampling in proportion would not provide enough cases to serve as an adequate basis for further analysis. To eliminate this difficulty, one may take a higher proportion of cases in the required stratum. Jahoda states “IN general terms, one can expect the greatest precision if the various strata are sampled proportionately to their variability’s with respect to the characteristic under study rather than proportionately in their relative sizes in the population”.

Stratification is a means to increase representativeness of the sample. It does not imply departure from the principle or randomness. The universe is divided into strata to adequately represent the various groups and then a random sample is selected form each stratum. The universe is divided into strata on the basis of research objectives i.e. criteria which are significant of research precision of results. According to Moser stratified sampling yields greater precision because. “The total variation in a population may be regarded a s composed of two elements variation between strata and variation within state. In stratified random sampling, variation between strata does not enter into the standard error at all, because one ensures that theist component of variation in the population is exactly reflected in the sample. There is no chance about it. Sampling, and there fore the play of chance only takes please within strata. Consequently since only the capitation within strata enters into the standard error, the greater the proportion of the total variation in a population that is accounted for by between-strata variation the greater will be the gain due to the strata that they differ as widely as possible from each other rand that the population within each stratum is as homogenous as it can be” Stratified random sampling contributes to more efficient being completely eliminated. All the strata are adequately represented and hence it yields more reliable results.

Great care must be taken while dividing the universe into strata. The definitions of the categories should be clear and explicit. Overlapping strata would yield defective sample.

The specifications must be carried out carefully. If there is a slip in carrying our plan, its theoretical advantages will be nullifies. If the investigators do not follow the instructions in selecting respondents, some selected cases may not be available, or may refuse to give interviews; the usual convenient procedure is to substitute them and to replace them. The sample actually obtained will not be a random sample.

Another significant point which must be kept in mind whole doing stratified sampling is that certain criteria provide in effective basis for stratification, but the relevant information may not be available at the time of drawing a sample but may become available only during the course of the survey. In such instance, if the criteria cannot be used in sampling, it should be applauding he analysis of the data.

For more help in Stratified Sampling click the button below to submit your homework assignment