22.214.171.124 Dealing with sampling error Generalising from a sample raises the issue of sampling error.
For example, the CASE STUDY sample shows a majority "against" making homosexuality illegal (see Table 126.96.36.199). Does this mean that the sampled population is also against illegalisation? We cannot jump to conclusions because we are dealing with a sample and the sample will not be an exact replica of the population that it was drawn from, even if it was an unbiased sample taken at random.
The variation in a sample that comes about from taking unbiased random samples is called sampling error and all samples have it. This is a nuisance but there is nothing that can be done to get rid of sampling error completely; it is part and parcel of sampling. The good thing about sampling error is that it is possible to estimate the extent of the error.
This makes sampling error very different from bias in samples. Bias arises when some members of the population do not have a chance of being in the sample, or where some members have a greater chance than others of being in the sample. Generalisations cannot really be made from highly biased samples.
Even if we assume the sample in our example survey was an unbiased random sample, it would still be possible to have selected a random sample who were particularly tolerant of homosexuality when the population, from which the sample was taken, is not tolerant. On the other hand, the sample may be particularly hostile to gays and lesbians. Or it may accurately reflect the population of 16 to 20-year-old college students in Birmingham. The problem is that we do not know which because we only have the information from the sample to go on. So what do we do? We estimate the probability that the variation between the sample mean (or percentage) and the population mean (or percentage) could be the result of sampling error. This is explained in Section 188.8.131.52.