~Biased Sampling~

~Characteristics of a Biased Sample:

~A sample taken from a population in such a way that it is prejudice in some way. Therefore, any conclusion drawn about the population is invalid, since it does not adequately represent the population as a whole.

~The logical structure of such a conclusion follows the following pattern:

~Concluding that the population, from which the sample is taken, has the attribute or statistic of our sample. This is called Hasty Generalization in logic.

~For example, There are 200 objects of different colors in a container. These objects are all the same size & shape, except 50 are red, 50 are green, 50 are blue, & 50 are black. A person randomly selects 40 for a sample. That sample, most likely, would contain objects of all 4 colors (to a high degree of probability) & would be a reasonable estimate of the proportion of colored objects in that container. However, if the red & green objects were the only ones that were magnetic & that person used a magnet to select the sample, there would only be red & green objects in the sample of 40, not a very good representation of the proportion of colors in that container (the population) since the blue & back objects will not be present..

~For example, if you wanted to know how people in the US feel about animal rights & decided to take a poll at a hunters convention, it would certainly be biased. I would confidently say, that most, if not all, would have a completely different view of killing animals for sport. So, taking a poll at a convention consisting of animal rights advocates, would produce completely different results.

~How to avoid Biased Sampling

~Basically, there are certain types of sampling, if done properly, that can avoid bias.

1) A Random Sample: A truly random sample taken from a population avoids bias since the major characteristic of  "randomness" is that every item in the population has an equal chance of being selected. That will assure a true representation of the population as a whole.

2) A Stratified Sample: For this type of sampling, the entire population is divided into subgroups (two or more)(strata). Each of these subgroups have a different attribute of the population but all members of each subgroup share that trait (i.e., separating the population being sampled into males & females, or separating the population into different income levels). Care must be taken that a random sample is taken from each stratum (subgroup) consistent with their size in the entire population. (i.e., if the population is estimated to consist of 5% high income people, 75% middle income people, & 20% of low income, then your sample should reflect this proportion as closely as possible in order for it to be truly representative of the population).