When collecting and analyzing data to form conclusions or propose potential causal relationships, it is important to take note of and state the biases that may be affecting the study's results. Continue reading to learn about them!
When trying to collect data on a population, it is often difficult to survey every unit in that entire population. Therefore, many statisticians, data scientists, businesses, and researchers collect data from samples, which are smaller sets of the population that can best represent it.
However, since not every individual can be accounted for in these sampling techniques, there are often sampling biases that cause them to have some statistical error. In this article, we will be specifically focusing on the five types of biases that are covered in AP Statistics.
Some types of biases are due to sampling error, or obtaining a sample that is not actually representative of the entire population that the data is intended to represent. These include undercoverage bias and voluntary response bias.
Undercoverage Bias
Undercoverage bias, also known as selection bias, is when the sample isn't a good representation of the entire population. This is usually a result of choosing participants by relying on a convenience sample, meaning that the group that participated in the study were selected from only one subgroup of the population with a certain common characteristic. For instance, if a person were to conduct a survey about what cars people own in a city and collects responses by going door-to-door in one neighborhood of the city to ask the house residents, this would be inadequate to represent the entire U.S. population, since people who have lower incomes who can't afford to live in that neighborhood would not be covered in the sample.
Voluntary Response Bias
Voluntary response bias happens when the individuals have the ability to select whether they want to be included in the sample or not. This method of sampling ends up overrepresenting those in the population who feel strongly about the topic, and underrepresent those who feel neutral or are indifferent. For instance, if all students were given an optional survey about the quality of the food served for school lunch, most likely, only students who strongly dislike or like the food will fill out the form, and those who feel indifferent or don't care will not. As a result, the survey will end up having more responses that are strongly positive and negative, and eliminate the true percentage of the population that don't have a strong opinion.
However, there are also some possible biases that can result from issues after the individuals of a sample are already chosen. These include nonresponse bias, response bias, and measurement error.
Nonresponse Bias
Nonresponse bias happens when individuals refuse to participate in a survey, or are unable to be found, after already being chosen to be part of the sample. This can influence the results of the survey if people who don't response have a common characteristic. For instance, if randomly selected customers of an electronic company were called to be asked about the quality of their phone, the people whose phones don't work properly will be unable to answer the call, causing them to not be adequately represented in the final results, and overestimate the quality of the phone.
Response Bias
Response bias happens when the responses collected from participants are inaccurate. This can be a result of inaccurate memory, biased questions, or social desirability of the participants. An example of a biased question would be "Is your favorite dessert served at the restaurant the bland cupcake, or everyone's favorite, the chocolate ice cream?" Posing a question in this manner influences certain opinions onto the responder that may affect their answer, and in this case, would pressure them to choose chocolate-flavored ice cream as the answer. Instead, it is good to make sure questions are as neutral as possible to avoid this occurrence, such as by simply saying "Is your favorite dessert served at the restaurant the cupcake or chocolate ice cream?" Furthermore, participants may lie in their responses due to social desirability, such as being hesitant to admit that they litter, since their response will be biased toward what they think other people would find more favorable to hear.
Measurement Error
Measurement error, also known as observational error, is the result of a mistake happening on the researcher's end, such as misrecording data or measurements. For instance, when collecting data on the heights of students in a class, the researcher could make an accidental typo for one of the data entries by entering "6 ft" instead of "5 ft", which could influence the overall result by stating that the class had a higher average height than it actually does.
Thank you for reading!
References:
“Survey Sampling Bias.” Stat Trek, stattrek.com/survey-research/survey-bias.aspx.
Comments