Bias means your sample does not represent the population about which you want to know something. Possible bias is the main reason it is difficult to generalize from a sample that was not selected randomly.

If your sample is not drawn from the population in which you are interested, the bias is known as non-selection bias. For example, if you choose a sample only from Internet users, you will not have information on people who do not use the Internet. When a sample is selected judgmentally, then the researcher decides which population member will be in the sample, and you may have selection bias. Generally, people like to talk with people whom they feel are like themselves, which means a judgmental sample is likely biased against inclusion of people unlike the researcher.

When people do not cooperate in a survey, there may be bias called non-response bias. People may not agree to be questioned in a mall, perhaps because they are too busy. Response bias happens when questions may be too difficult ("How many shirts did you buy in the past five years?"), too sensitive ("What is your annual income?"), or too embarrassing ("How many hours a day do you watch game show reruns?") The best way to handle bias is simply to think through your research plan very carefully and hope that you have identified all the possible sources of bias.

The best way to handle bias is simply to think through your research plan very carefully and hope that you have identified all the possible sources of bias.

Random Errors

Whenever you use a sample, there are random errors. Random error occurs because the composition of each sample will likely be somewhat different, leading to somewhat different results. The variation in the results from different samples is measured by random error. The smaller the random error, the more accurate the information from your sample.

You make random error smaller by increasing the size of your sample. However, you should keep in mind that random error does not decrease with the sample size; it decreases with the square root of the sample size. That means that to decrease a random error from, say, 0.50 cans of soft drink to 0.25 cans of soft drink—a decrease of 50 percent, you need to quadruple the size of the sample, from, say, 100 to 400 respondents.

Sample Size

Determining the optimal sample size for a study can be very complicated. However, you can use some rules-of-thumb to help you decide on your sample size.

The minimum sample size for any population such as a target market segment should be 30. The reason that 30 is somewhat of a "magic number" is that with sample sizes of 30, most statisticians would allow the use of the well-known normal distribution as an approximation when doing analysis even if your target population is not "normal." With data from sample sizes below 30, unless your target population is itself normally distributed, you may need to use approaches other than the normal distribution to make estimates from your data.

If you can afford it, you would like a sample size larger than 30. Generally, you want a larger sample size:

• When the target population is heterogeneous. The more varied your customers, the larger sample size you need. If all your customers are the same or similar, then a smaller sample is sufficient. For example, if you are cooking soup and you stir it well, how many spoonfuls do you need to see how it tastes? Likely just one.

• When the members of a population are very important. It is more important for you to survey more large customers than small customers because a large customer 's actions will have more of an impact on your sales.

Was this article helpful?

0 0

Post a comment