1.9 Sampling

How can we find out what people in Finland think about nuclear power? It would be far too laborious to try and contact every person in the country and ask them their opinion. It is better to take a sample.

A sample is a smaller group selected from the target group in order to draw a picture of certain characteristics of that group. We call this group the statistical population. Finnish people of working age are often taken as the population of opinion polls, for example.

Samples are usually selected by some random device, giving what is known as a random sample.

One alternative to randomization is the method of systematic sampling. This can be done from a list where individuals are placed in a random order (alphabetical order will do because that is random as far as people's opinions are concerned) and selecting the sample by taking sampling units at equally spaced intervals. It is also easy to pick out telephone numbers at random to obtain a random sample of respondents (although this excludes respondents who do not have a phone). One way to draw a random sample of the Finnish population is to list the whole population in alphabetical order or in order of their place of residence and then select every five thousandth name from this list; this will give a random sample of around 1,000 persons.

The size of the sample is important

If a sample is drawn randomly, it is in principle representative of the population. Random factors come into play in every individual observation, but when there is a sufficiently large number of chance occurrences, they cancel out one another.

The key consideration with respect to sample size is the level of detail aimed at in the analysis. If the researcher wants to divide the sample into various subgroups and say something reliable about them, it will be necessary to increase the sample size. The size of the population, on the other hand, is as such irrelevant to the sample size. In other words, the opinions of American citizens can in principle be studied with the same size of interview sample as the opinions of Finnish citizens.

A sample of 1,000 persons provides a reasonably reliable basis for describing the population if the phenomena concerned are simple and concern the whole population. The agreed principle in the statistics field is that this sample size allows results with an accuracy of a few percentage points. The margin of error increases rapidly to over ten percentage points when the sample size decreases.

If the results are examined by different age groups, for example, the margin of error has to be calculated separately for each age group.

Next lesson: The basic concepts of social and economic statistics