SOURCES OF DATA

In stratified sampling, there is some sort of numerical order that can be seen in the data. For example, you could sample the population for the occurrence of an illness and you could then stratify by age eg 100 people under 30, 200 30 – 60 and 50 over 60. Sometimes the size of the sample in each stratum will reflect the size of the population in the stratum.

It is commonly done in auditing eg to test receivable balances where 10,000 accounts owed <$1,000 each and 5 owe more than $100,000, you might say that all accounts over $10,000 will be examined plus 100 of the small ones. Pure random sampling could easily miss all the large balances.

Multistage sampling does not have a natural order. So if you wanted to survey 5,000 people randomly chosen in the population it would be very expensive travelling all over the place to find the ones selected randomly eg by Identity card number. Instead the sample could be chosen first by randomly choosing 5 counties, then by choosing 10 post codes in each then picking 100 people at random in each post code. This is a more efficient way to collect data but, once pure random sampling is not used, great care is needed in drawing conclusions form the data.

No problems.