Exercise 40 Page 759

We are given the information that a toothpaste company, based on survey results, is claiming that four out of five dentists recommend their toothpaste. Before it is possible to determine whether the results are unbiased, we need to know the following facts about the survey.

• Sampling method
• Sample size
• Survey question

  Since the company is reporting about dentists in general, the most likely population is all dentists for whom the toothpaste is available. For the results to be unbiased, a sample selected from the population must be unbiased. This means that it must be representative of the population. Let's consider the following factors that could introduce bias to the sample.

  Factor	Potential Bias
  Sampling method	If the survey was optional, dentists who chose to respond are more likely to be ones who would recommend the toothpaste. A self-selected is usually biased.
  	A survey conducted in the city where the headquarters of the company is located would not give valid results. A convenience sample is usually biased.
  	Dentists who are paid to promote the toothpaste are less likely to give objective answers to the survey.
  Sample size	It is important that the sample size is large enough compared to the number of dentists who may have tested the toothpaste. Statisticians claim that the size should be more than 10 % of the size of the population.
  Location of the survey	A survey conducted in only one city or only one state will not give valid results. The location of the survey should be adjusted to the availability of the toothpaste.
  The survey was conducted by the manufacturer.	A survey conducted by an interested party can lead to confirmation bias. This means that, because the company wants or expects their toothpaste to be recommended, their desires could influence the data analysis. An independent company should conduct the survey.

  Next, let's consider the survey question. A survey question is biased when it contains assumptions that may or may not be true. For example, the question may assume that a dentist has used the toothpaste which may not be true. Additionally, let's consider the following biased survey question.

  Biased Survey Question

  Would you recommend our toothpaste over other cheap and low-quality toothpastes?

  In this example question, the highlighted part can make one answer seem better than the other. Also, the words cheap and low-quality may evoke strong feelings in some dentists which may lead them to give what they feel is the correct answer. Therefore, we also need to know the information on the survey question to determine whether the results are unbiased.