Brittany Caplin

3/2/10

In the Austin and Pinkleton chapter they discuss the basics of a sample, which is a relatively small number of target audience members. There are other important definitions like: a population, which consists of all the members in the group, a census which is collecting information on the entire population, and parameters which are a characteristic or property of a population. After data is collected from the sample, an inference is made by experts to figure out how that information applies to the rest of the population. There are two types of sampling: probability and non-probability. Probability sampling is random so each member has a chance of being selected, whereas in non-probability sampling, we do not know the chances of each member being selected and therefore do not know if it’s equal. The next thing that Austin and Pinkleton focus on is sampling error, which is the difference of opinions and attitudes between the sample and the population. Standard deviation is how to measure the error and a sampling distribution the grouping or arrangement of characteristics that the researchers want to measure. The frequency of the characteristics are recorded on a measurement scale and analyzed. Standard deviation is the tool used to measure error and the area beneath a normal curve can reflect the margin of error. Confidence levels and intervals express a researcher’s margin of error in a percentage form. It usually is around 3-5% confidence level in order to be considered a well done experiment.

According to Gawiser and Witt, sampling error is not only that we made a mistake but “the poll results may not match the true opinions of the entire population exactly.” Sampling error is very easy to the most common because it is simple to calculate the mistake and easy to report. Another definition that Gawiser and Witt provide for sampling error is the amount of chance variation expected in a series of samples of the population. There are two main aspects that are important for the variation- the size of the sample and the size of the population. The size of the sample is the key variable in the calculation of sampling error. Sample errors are found in every study and as a journalist or PR specialist, I need to know how to decipher and report them. The general population does not understand the significance of a sampling error so it’s important to put it into other terms. The chapter also warned that sampling errors are not the only things wrong with a poll and other errors within sampling process should be examined. In the Appendix, there were examples of how to calculate probability, for example of a coin, and what it means. It was also deemed a crash course in basic statistics and gave examples of the calculations needed to determine sample error and confidence intervals.

Prior to this class I didn’t know how to interpret sample of error. I took a year of stats but sample of error is more than just mechanically doing an equation. This reading helped me understand that sampling error is important in my future career as a PR specialist because I’m going to need to be able to detect and interpret the error. I cannot take a poll at face value because it could be misleading and cause me to make a mistake in my career. I really appreciated how the book explained these statistic terms instead of just giving equations, which is how my high school class operated. I know have a clearer understanding that no poll is perfect but it is important to look at the sample size versus the population size and to determine if they have an acceptable level of confidence for the poll. Another important part of this chapter is how to report your findings. The common public will not understand if in an article for a newspaper you state that the confidence interval is 95%. However, to explain it in terms referencing the percentage out of 20 people or 100 people can be easier to understand. I don’t have a lot of math knowledge but I’m encouraged to take a statics class to refresh my memory.

I found an article from Pollster that deals with the issues of sampling error from a previous poll from SurveyUSA. A survey was taken about which candidate, Obama or McCain, would win in each state. The sample size was only 600 per state, which cause a controversy in this poll. For the population of registered voters per state, a sample size of 600 was not representative enough for the voters opinions on the candidates. SurveyUSA also had another underlining problem that it weighed its data, however this was not taken into account by Pollster. Another issue was they only polled previous voters, not likely or potential voters. However, the main problem in the article was the confidence level. SurveyUSA did not specify what their confidence levels were but Pollster figured it out and they were inconsistent. For a state to have a strong candidate preference they needed a 95% confidence level. Then a “lean” candidate preference has a 68% confidence level, and the other states were toss ups. The problem with the SurveyUSA study is that they categorized all of the “lean” states as preferential for a candidate. But, a 68% confidence level isn’t a strong enough level for a 600 person sample against a population of potentially hundreds of thousands of people. Therefore, the poll was misleading because some of the states shouldn’t have been categorized for Obama or McCain. This is a perfect example of how voting or presidential polls are often the most misleading poll and commonly the ones with the most mistakes.

## Leave a Reply