Stratified Random Sampling:

This method of sampling is used when the population is comprised of natural subdivision of units, The method consist in classifying the population units into a certain numbers of groups called strata. Then random samples are selected   independently from each group. The division of the population into strata is done in such a way that.

(1)   There is greater homogeneity within each group also  called stratum;

(2)   There is no over lapping in various strata or groups.

In practice geographical, sociological and economic characteristics are used for dividing the universe into different strata.  For example, if is desired to study the wage levels of workers in an industry, it would be better to classify all works into skilled and unskilled males and females categories and then select random simple   from each group  in the ratio to the size of that group  in population. The size of sample from each strata may either be proportional or disproportional to  the size of each.

(a)    Proportional Size: Under this a given percentage of items are selected from each stratum. For example in case of 50% selection there will be a selection of 5 to 10 units and 10 of20 units. Consisting of 10 and 20 units. This selection will be at random following any of the methods discussed earlier.

(b)   Disproportional Size: A disproportional selection may also be followed if different strata reveal different levels of within strata variability. Following the normal sampling rules, a larger sample is recommended if some sub strata reveal greater degree of variability with regard to a phenomenon relevant to the inquiry in comparison to the general lot.