Quantitative Reasoning Preview Assignment 15.A Preview Assignment 15.A Preparing for the next class In the next class, you will learn more about experimental studies. Remember that, in experimental studies, researchers manipulate an explanatory variable to observe changes in a response variable. Learn more about variables at http://www.oswego.edu/~srp/stats/variable_types.htm. If you find some variable types with which you are unfamiliar, add them to your personal statistics dictionary. 1) Read the passage about quantitative and categorical variables. Part A: Which of the following is not an example of a quantitative variable? a) b) c) d) Height Political affiliation Temperature Body Mass Index Part B: Which of the following is not an example of a categorical variable? a) b) c) d) Hair color Gender Field of study Grade point average 2) Researchers are designing an experiment to learn whether eating dark chocolate can help people have healthy hearts. They plan to give various doses of dark chocolate to patients for a long-term study. After a specified amount of time, they will compare heart health among the patients. Part A: What is the explanatory variable in this study? a) The patients b) The chocolate dose c) The heart health of patients The Charles A. Dana Center at The University of Texas at Austin 1 Version 1.0 Quantitative Reasoning Preview Assignment 15.A Part B: What is the response variable? a) The patients b) The chocolate dose c) The heart health of patients Part C: What is the manipulated treatment in this study? a) The patients b) The various doses of chocolate c) The various levels of heart health Monitoring your readiness 3) To effectively plan and use your time wisely, it helps to think about what you know and do not know. For each of the following, rate how confident you are that you can successfully do that skill. Use the following descriptions to rate yourself: 5—I am extremely confident I can do this task. 4—I am somewhat confident I can do this task. 3—I am not sure how confident I am. 2—I am not very confident I can do this task. 1—I am definitely not confident I can do this task. Skills needed for Lesson 15, Part A Skill or Concept: I can . . . Question used for understanding Distinguish between quantitative and categorical variables. 1 Distinguish between explanatory and response variables. 2 Identify treatment levels. 2 Rating from 1 to 5 Now use the ratings to get ready for your next lesson. If your rating is a 3 or below, you should get help with the material before class. Remember, your instructor is going to assume that you are confident with the material and will not take class time to answer questions about it. The Charles A. Dana Center at The University of Texas at Austin 2 Version 1.0 Quantitative Reasoning Student Pages 15.A, The video game diet Lesson 15, Part A The Video Game Diet Consider the following research study. Research question: Dr. Elizabeth Jackson, a professor at the University of Michigan Health System, surveyed 1,003 middle-schoolers about their snacking habits and their predominant after-school activity (outdoor activities, video games, or watching TV).1 Conclusion: Kids who spend most of their time outdoors eat the fewest calories from unhealthy snacks. (Interestingly, kids who played the most video games ate fewer calories from unhealthy snacks than those who tended to watch more TV.) 1) In a statistical study, a variable is an observation that changes from individual to individual. Part A: What are the two variables in this study? Credit: ThinkStock Objectives for the lesson You will understand that: o In experimental studies, researchers manipulate an explanatory variable to observe the effects on a response variable. o To manipulate the explanatory variable, researchers divide subjects into similar groups and apply a different treatment to each group. o If the groups are truly similar, then differences observed in the response variable might be attributed to the explanatory variable. You will be able to: o Determine when a study design allows a conclusion to be made about cause and effect. o Design an experimental study. Part B: Identify the explanatory and response variables in the University of Michigan study and classify them as categorical or quantitative. Explanatory variable: Response variable: 1 Singh, M. (2014, March 28). Why playing Minecraft might be more healthful for kids than TV. Retrieved November 23, 2014, from http://www.npr.org/blogs/health/2014/03/28/295692162/why-playing-minecraftmight-be-healthier-for-kids-than-tv. The Charles A. Dana Center at The University of Texas at Austin 3 Version 1.0 Quantitative Reasoning Student Pages 15.A, The video game diet Part C: Can researchers in the study conclude that the amount of after-school activity was the cause for the number of calories that kids consumed from unhealthy snacks? Can you offer a possible alternate explanation? In an observational study, researchers simply observe the variables of interest. In an experimental study, researchers actively manipulate the explanatory variable. Part D: Was this an observational or experimental study? Justify your answer. Part E: In what type of study can we make conclusions about cause and effect? 2) Let’s redesign the study about middle-schoolers’ after-school activities into an experimental study to see if a cause-and-effect conclusion can be made. Part A: The explanatory variable is after-school activity. In an experimental study, should the kids (participants) pick an activity according to their preference, or should researchers control this? Part B: What does it mean in this context to say that “researchers manipulate the explanatory variable in an experiment?” Part C: In an experimental study, the values of the explanatory variable are called treatments. What are the treatments in this study? Part D: Suppose you decide to apply the treatments to different groups by dividing children randomly among three groups. Would there be any reason to think that one group would be very different from another? Part E: In an experiment, random assignment into groups tends to create groups that are similar. The only real difference from group to group is the explanatory variable that researchers are manipulating. If the children assigned to the video game group consumed fewer calories than the children assigned to the TV group, could there be another explanation other than the activity? Part F: If the children in the video game group consumed fewer calories than the children in the TV group, what conclusion could be made about the activity? In experiments, if we apply the various treatments to groups that are similar in composition, then we can conclude that they are responsible for any differences observed in the response variable. The Charles A. Dana Center at The University of Texas at Austin 4 Version 1.0 Quantitative Reasoning Suggested Instructor Notes 15.A, The video game diet Lesson 15, Part A The Video Game Diet Overview and student objectives Overview Lesson Length: 25 minutes In this lesson, we begin to look closely at experimental studies. Some of the ideas in the lesson were discussed briefly in earlier lessons. We review those concepts and continue to develop them here. The lesson focuses on explanatory and response variables, and the idea that in experiments, researchers manipulate an explanatory variable to observe changes in a response variable. Prior Lesson: Lesson 14, Part C, “Conclusions in Observational Studies” Objectives Next Lesson: Lesson 15, Part B, “All Things in Moderation” (25 minutes) Constructive Perseverance Level: 1 Theme: Health Literacy Students will understand that: Goals: Reasoning, Evaluation • In experimental studies, researchers manipulate an explanatory variable to observe the effects on a response variable. • To manipulate the explanatory variable, researchers Outcomes: N1, C1, C2 divide subjects into similar groups and apply a different treatment to each group. • If the groups are truly similar, then differences observed in the response variable might be attributed to the explanatory variable. Quantitative Reasoning Outcomes: S1, S2, S3, Related Foundations Students will be able to: • Determine when a study design allows us to make a conclusion to be made about cause and effect. • Design an experimental study. Suggested resources and preparation Materials and technology • Computer, projector, document camera • Preview Assignment 15.A • Student Pages for Lesson 15, Part A • Practice Assignment 15.A Prerequisite assumptions Students should be able to utilize the basic vocabulary related to the design of statistical studies. The Charles A. Dana Center at The University of Texas at Austin 5 Version 1.0 Quantitative Reasoning Suggested Instructor Notes 15.A, The video game diet Making connections This lesson: • Connects back to Lessons 3 and 13 activities about samples, populations, and numerical summaries. • Connects forward to more lessons on statistical studies (Lessons 19–21). Background context In Preview Assignment 15.A, students were given instructions to learn more about variables at http://www.oswego.edu/~srp/stats/variable_types.htm. Suggested instructional plan Frame the lesson (6 minutes) ThinkPair-Share • Ask students to read the introduction to the research study and answer question 1, Part A. • Take responses from students and then ask the class to refine the responses. For example, students may say “activity” and “snacks.” Ask: o “What about the activities? Is the variable the amount of time spent performing an activity?” [Answer: No, it is the type of activity.] o “What about snacks? Is it the number of snacks? The type of snack?” [Answer: No, it is the amount of calories consumed.] • These refinements will help with determining whether the variables are categorical or quantitative, and with writing a clear conclusion. • Also, writing clear variable descriptors can help students to choose an appropriate graphical display when asked to do so. • Transition to the lesson activities by briefly discussing the Objectives for the lesson. Lesson activities (15 minutes) Group Work then Whole Class Discussion • Ask students to complete question 1. Hold a whole class discussion to ensure that everyone is clear on the vocabulary associated with the example. • Students often confuse the explanatory and response variables. Remind them that the explanatory variable is the one that may (but may not!) influence changes in the other variable. o “Do we think the activity might explain the calories, or that the calories explain the activity?” The Charles A. Dana Center at The University of Texas at Austin 6 Version 1.0 Quantitative Reasoning Suggested Instructor Notes 15.A, The video game diet o “Do calories respond to activity, or does activity respond to calories?” • Group Work This study is observational in nature, and as students have learned, observational studies do not allow cause and effect conclusions. It may be that kids who watch more TV are more complacent physically and more inclined to eat snacks. o “Is it possible that kids who watch a lot of TV have traits in common with kids who snack a lot? What might those traits be?” Question 2 • If students have trouble with the idea of manipulating a variable, remind them that manipulating variables means that the researcher controls the amount or type of variable that is applied to the participant. In such an experiment, we would not just observe who watches a lot of TV; we would actually control the amount each participant watches. • Students should apply the idea of manipulating a variable to this context. In this case, students are told what their activity will be. If we let students pick, then the study is not an experiment. o “Who should control the explanatory variable—the students or the researchers?” • We are attempting to help students understand how the abstract terms such as variable and value apply in this context. It may be difficult at first, so remind them that treatments are the manipulations of the explanatory variable. The explanatory variable is activity, so how is this being manipulated? o “If the explanatory variable is an activity, then what are the possible values (or outcomes) of this variable?” • Because the groups are presumably similar, there is only one thing that varies from group to group—the value of the explanatory variable. Since it is the only variable, it is the only thing to explain any responses observed. The main idea is that we are limiting variation between the groups except for the variable that we are manipulating. If the groups are different in only one regard, then that difference is responsible for changes observed. • Researchers use various measures to ensure that the groups are similar or to what degree they are not. o “Pretend the groups are exactly the same except in their activity after school. If this is the only difference, then is there any other explanation for the changes in calories consumed?” • Since it is appropriate to make a causal conclusion, students can say that the values of the explanatory variable are the causes for differences in the groups. Just as you discouraged the use of that word in conclusions of observational studies, encourage students to use the word causes in their conclusions of experimental studies. Guiding Questions The Charles A. Dana Center at The University of Texas at Austin 7 Version 1.0 Quantitative Reasoning Suggested Instructor Notes 15.A, The video game diet Wrap-up/transition (4 minutes) Wrap-up • Review the main ideas of the lesson and remind students to complete their personal statistics dictionary. o In experimental studies, researchers are concerned about explanatory variables and response variables. o To make a cause-and-effect conclusion, researchers must divide their subjects into groups that are similar. o Researchers manipulate the explanatory variable by assigning a quantity of the variable to each of the similar groups. o If changes are observed in the response variable, then we can assume that the values of the explanatory variable are the cause. • Have students refer back to the Objectives for the lesson and check the ones they recognize from the activity. Alternatively, they may check objectives throughout the lesson. • “For the remainder of Lesson 15, we will continue to build on the concepts of experimental studies.” Transition Suggested assessment, assignments, and reflections • Give Practice Assignment 15.A. • Give the Preview Assignments, if any, for the lesson activities that you plan to complete in the next class meeting. The Charles A. Dana Center at The University of Texas at Austin 8 Version 1.0 Quantitative Reasoning Suggested Instructor Notes 15.A, The video game diet Lesson 15, Part A The Video Game Diet – ANSWERS Consider the following research study. Research question: Dr. Elizabeth Jackson, a professor at the University of Michigan Health System, surveyed 1,003 middle-schoolers about their snacking habits and their predominant after-school activity (outdoor activities, video games, or watching TV).1 Conclusion: Kids who spend most of their time outdoors eat the fewest calories from unhealthy snacks. (Interestingly, kids who played the most video games ate fewer calories from unhealthy snacks than those who tended to watch more TV.) 1) In a statistical study, a variable is an observation that changes from individual to individual. Part A: What are the two variables in this study? Credit: ThinkStock Answer: The type of after-school activity and the amount of calories from unhealthy snacks. Objectives for the lesson You will understand that: o In experimental studies, researchers manipulate an explanatory variable to observe the effects on a response variable. o To manipulate the explanatory variable, researchers divide subjects into similar groups and apply a different treatment to each group. o If the groups are truly similar, then differences observed in the response variable might be attributed to the explanatory variable. You will be able to: o Determine when a study design allows a conclusion to be made about cause and effect. o Design an experimental study. 1 Singh, M. (2014, March 28). Why playing Minecraft might be more healthful for kids than TV. Retrieved November 23, 2014, from http://www.npr.org/blogs/health/2014/03/28/295692162/why-playing-minecraftmight-be-healthier-for-kids-than-tv. The Charles A. Dana Center at The University of Texas at Austin 9 Version 1.0 Quantitative Reasoning Suggested Instructor Notes 15.A, The video game diet Part B: Identify the explanatory and response variables in the University of Michigan study and classify them as categorical or quantitative. Explanatory variable: Answer: Type of after school-activity; categorical. Response variable: Answer: Amount of calories consumed from unhealthy snacks; quantitative. Part C: Can researchers in the study conclude that the amount of after-school activity was the cause for the number of calories that kids consumed from unhealthy snacks? Can you offer a possible alternate explanation? Answer: No. It may be that children who chose to watch a lot of TV had things in common with kids who ate more calories in snacks. Maybe there was less parental intervention that caused both of these variables to increase together. In an observational study, researchers simply observe the variables of interest. In an experimental study, researchers actively manipulate the explanatory variable. Part D: Was this an observational or experimental study? Justify your answer. Answer: This was an observational study. No attempt was made to manipulate the after school activity. Part E: In what type of study can we make conclusions about cause and effect? Answer: Only in (well-designed) experimental studies. 2) Let’s redesign the study about middle-schoolers’ after-school activities into an experimental study to see if a cause-and-effect conclusion can be made. Part A: The explanatory variable is after-school activity. In an experimental study, should the kids (participants) pick an activity according to their preference, or should researchers control this? Answer: The researchers should control the explanatory variable. The Charles A. Dana Center at The University of Texas at Austin 10 Version 1.0 Quantitative Reasoning Suggested Instructor Notes 15.A, The video game diet Part B: What does it mean in this context to say that “researchers manipulate the explanatory variable in an experiment?” Answer: It means that the researchers will determine each student’s activity. Part C: In an experimental study, the values of the explanatory variable are called treatments. What are the treatments in this study? Answer: Outdoor activities, playing video games, and watching TV. Part D: Suppose you decide to apply the treatments to different groups by dividing children randomly among three groups. Would there be any reason to think that one group would be very different from another? Answer: If the groups are created randomly, there is no reason that they would be very different. Part E: In an experiment, random assignment into groups tends to create groups that are similar. The only real difference from group to group is the explanatory variable that researchers are manipulating. If the children assigned to the video game group consumed fewer calories than the children assigned to the TV group, could there be another explanation other than the activity? Answer: There is no other explanation, other than chance variation in the random assignment into groups. Part F: If the children in the video game group consumed fewer calories than the children in the TV group, what conclusion could be made about the activity? Answer: We can conclude that the activity was the cause of the differences in calories from unhealthy snacks. In experiments, if we apply the various treatments to groups that are similar in composition, then we can conclude that they are responsible for any differences observed in the response variable. The Charles A. Dana Center at The University of Texas at Austin 11 Version 1.0 Quantitative Reasoning Suggested Instructor Notes 15.A, The video game diet The Charles A. Dana Center at The University of Texas at Austin 12 Version 1.0 Quantitative Reasoning Practice Assignment 15.A Practice Assignment 15.A In a previous practice assignment, you learned about a study in which volunteers (who were all around 30 years of age) from the Chicago area wore a light sensor to measure their exposure to sunshine.1 It was found that those who experienced more daylight had lower body mass indexes (BMIs). The study was observational. It was not an experiment, so conclusions about cause and effect could not be made. 1) Let’s redesign the observational study into an experimental study. Part A: What is the explanatory variable for the new study? a) Body mass index b) Sunshine exposure c) Age Part B: What is the response variable in the new study? a) Body mass index b) Sunshine exposure c) Age Part C: In experiments, we divide people into groups and apply treatments to each group. How should we create these groups in the new study? a) By grouping people according to their BMI. b) By grouping people by how much time they spend in the daylight. c) By randomly assigning people to groups. Part D: As we apply treatments to the groups, we manipulate the explanatory variable. How should this be done in the new study? a) By assigning to each group a specific number of hours each day that they are to spend in the daylight. b) By giving each member a light sensor and observing how much time each group spends in daylight each day. c) By assigning the groups with higher average BMI more time in the daylight so that we can see if they lose weight. 1 Aubrey, A. (2014, April 3). Good day sunshine: Could morning light help keep us lean? Retrieved November 19, 2014, http://www.npr.org/blogs/health/2014/04/03/298358419/good-day-sunshine-couldmorning-light-help-keep-us-lean. The Charles A. Dana Center at The University of Texas at Austin 13 Version 1.0 Quantitative Reasoning Practice Assignment 15.A Part E: After a period of time, suppose that there is no difference in average BMI from group to group. What would be the best conclusion? a) There is evidence of an association between amount of sun exposure and BMI. b) There is evidence that increased sun exposure contributes to lower BMI. c) There is no evidence that increased sun exposure contributes to lower BMI. Part F: After a period of time, suppose that groups who received the most sunlight had the lowest average BMI. What would be the best conclusion? a) There is evidence of an association between amount of sun exposure and BMI. b) There is evidence that increased sun exposure contributes to lower BMI. c) There is no evidence that increased sun exposure contributes to lower BMI. 2) Research the Carolina Abecedarian Project website: abc.fpg.unc.edu. Use the left menu bar to read “About the Project,” “Major Findings,” and “Policy Implications.” Part A: The project was an ___________ study. a) Experimental b) Observational Part B: How were the children organized into groups? a) b) c) d) By socioeconomic class: lower, middle, and upper-class groups. By race/ethnicity. By gender: males in one group, females in another. By random assignment. Part C: If the randomization process created the groups to be similar enough, we can infer that the treatment is the cause. Under this assumption, make an inference about the treatment’s effect relating to college-going behaviors. a) There is evidence that the intervention contributes to a higher likelihood of attending college. b) There is evidence that attending college contributes to a higher likelihood of being in the intervention group. The Charles A. Dana Center at The University of Texas at Austin 14 Version 1.0 Quantitative Reasoning Practice Assignment 15.A Part D: The sample used in the Carolina Abecedarian Project is not representative of the entire population of American infants. Still, there may be a sampling frame—a subset of the population that the sample does represent. Because the participants in the study were born in the mid-1970s and were from lowincome families, the sampling frame or subset represented by the study is low-income children in the mid-1970s. (Although not mentioned on this website, they were also mostly African American). Experiments focus on making inferences about treatments, but treatments are applied to populations. Would it be reasonable to infer that all children who attend quality preschool would receive benefits similar to those observed in the study? a) Yes b) No The Charles A. Dana Center at The University of Texas at Austin 15 Version 1.0 Quantitative Reasoning Practice Assignment 15.A The Charles A. Dana Center at The University of Texas at Austin 16 Version 1.0 Quantitative Reasoning Preview Assignment 15.B Preview Assignment 15.B Preparing for the next class The next class will focus specifically on experimental studies. You will need to distinguish an observational study from an experimental study. 1) Consider the following studies. Study I: A researcher randomly divided 200 high school students into two groups. The first group was told to not exercise. The second group was told to exercise at least 5 hours each week. At the end of two months, the heart rates of each participant in each group were measured at rest. It was found that the group that exercised 5 hours a week had a significantly lower average heart rate. Study II: A researcher gathered 200 high school students randomly and asked them how long they exercised each week and then measured their heart rates at rest. It was found that students who exercised at least 5 hours a week had a significantly lower average heart rate. Part A: Which study was experimental? Why? a) Study I b) Study II Part B: List the explanatory and response variables in the studies. a) Explanatory: heart rate. Response: exercise time. b) Explanatory: exercise time. Response: heart rate. c) Explanatory: student. Response: amount of exercise. Part C: State whether the variables in the studies are quantitative or categorical/ qualitative. a) b) c) d) Explanatory: categorical. Response: quantitative. Explanatory: categorical. Response: categorical. Explanatory: quantitative. Response: quantitative. Explanatory: quantitative. Response: categorical. The Charles A. Dana Center at The University of Texas at Austin 17 Version 1.0 Quantitative Reasoning Preview Assignment 15.B Part D: Identify a flaw in Study I above. a) The participants should not have been randomly assigned to groups. b) The participants should not have been told how much to exercise. c) The participants’ heart rates should have been measured both before and after the study. 2) In statistics, sometimes a “confounding variable” appears in a statistical model. Search for information and examples of confounding variables. Write a definition in your own words and give two brief examples. Be sure to give your source. Monitoring your readiness 3) To effectively plan and use your time wisely, it helps to think about what you know and do not know. For each of the following, rate how confident you are that you can successfully do that skill. Use the following descriptions to rate yourself: 5—I am extremely confident I can do this task. 4—I am somewhat confident I can do this task. 3—I am not sure how confident I am. 2—I am not very confident I can do this task. 1—I am definitely not confident I can do this task. The Charles A. Dana Center at The University of Texas at Austin 18 Version 1.0 Quantitative Reasoning Preview Assignment 15.B Skills needed for Lesson 15, Part B Skill or Concept: I can . . . Question used for understanding Distinguish observational from experimental studies. 1, Part A Determine explanatory and response variables. 1, Part B Distinguish quantitative and categorical variables. 1, Part C Identify a flaw in an experimental study design. 1, Part D Define a confounding variable. Rating from 1 to 5 2 Now use the ratings to get ready for your next lesson. If your rating is a 3 or below, you should get help with the material before class. Remember, your instructor is going to assume that you are confident with the material and will not take class time to answer questions about it. Ways to get help: • see your instructor before class for help • ask your instructor for on-campus resources • set up a study group with classmates so you can help each other • work with a tutor The Charles A. Dana Center at The University of Texas at Austin 19 Version 1.0 Quantitative Reasoning Preview Assignment 15.B The Charles A. Dana Center at The University of Texas at Austin 20 Version 1.0 Statistical Reasoning Student Pages 15.B, All things in moderation Lesson 15, Part B All Things in Moderation Chocolate has been a favorite food of humans for over 3,000 years, originating from Central America and Mexico, where it was enjoyed as a drink. It was thought to have certain medicinal properties and was even considered valuable enough to be used as money.1 Lately, studies have emerged that also promote the health benefits of chocolate. 1) Have you heard of any health benefits from eating chocolate? If so, what were they? Credit: ThinkStock Objectives for the lesson You will understand that: o Conclusions about cause and effect cannot be made in observational studies because of confounding variables. o Conclusions about cause and effect can be made in experimental studies where confounding is controlled. o Experimental groups created through random assignment can usually control variability among confounding variables. You will begin to be able to: o Analyze a statistical study and identify possible confounding variables. o Decide when confounding variables restrict conclusions about cause and effect. o Design an experiment that allows conclusions about cause and effect. In this lesson, we consider how conclusions in statistical studies can be uncertain when researchers do not know why a particular outcome has happened. 2) A research study in Sweden asked the research question: “Does chocolate reduce the risk of stroke?”2 1 Benson, A. (2008, March 1.) A brief history of chocolate. Retrieved November 24, 2014, from http://www.smithsonianmag.com/arts-culture/a-brief-history-of-chocolate-21860917/?no-ist. 2 Larsson, S. C., Virtamo, J., & Wolk, A. (2012). Chocolate consumption and risk of stroke: A prospective cohort of men and meta-analysis. Neurology, 79(12), 1223-–1229. The Charles A. Dana Center at The University of Texas at Austin 21 Version 1.0 Statistical Reasoning Student Pages 15.B, All things in moderation A sample of 37,000 Swedish men were divided into four groups based on their normal chocolate consumption, ranging from those who ate no chocolate to the group who ate the most (63 grams per week). It is important that the study included so many people because it makes the results more reliable. Participants were followed for 10 years. The study concluded that men who ate more chocolate had a lower risk of experiencing a stroke. Part A: Did this study prove that eating chocolate causes a lower risk of stroke? Explain your response. Part B: Offer an alternate explanation for this lowered risk among chocolate eaters. 3) In Preview Assignment 15.B, you read about confounding variables, or factors that give alternate explanations for an effect. When there are confounding variables, researchers cannot make inferences about cause and effect. Another article3 pointed out some important facts in the study described above. The men who ate the most chocolate tended to be younger, were more educated, and were less likely to smoke or high blood pressure. They also reported eating more vegetables and drinking more wine. All of these factors have positive health benefits. Part A: Explain how this information sheds light on the study that says that eating chocolate lowers the risk for stroke? Does it make us more or less certain about the effects of eating chocolate? Part B: Chocolate eaters tended to be more highly educated. What other confounding variables arise from this fact that may also cause a lowered risk of stroke? (What other things do people who are better educated tend to have in common?) Cause and effect conclusions cannot be made in observational studies because of confounding variables. Experiments allow such conclusions by holding confounding variables constant among several groups. To do this, researchers make sure that groups are similar in terms of education, age, eating habits, and so on. When groups are similar, confounding variables are controlled (not varying), and it becomes easier to argue that a treatment (like chocolate) is the cause of an observed effect (like fewer people having a stroke). 3 Goodman, B. (2013, January 7). Don’t fudge the facts on chocolate studies. Retrieved November 24, 2014, from http://healthjournalism.org/blog/2013/01/dont-fudge-the-facts-on-chocolate-studies. The Charles A. Dana Center at The University of Texas at Austin 22 Version 1.0 Statistical Reasoning Student Pages 15.B, All things in moderation To control for the confounding variables, researchers divide participants into similar groups. Some groups are treatment groups—they receive a real treatment. Other groups are control groups—they receive a placebo (a fake treatment) or nothing at all. Control groups are used as a basis for comparison to the treatment groups. 4) We want to design a new study to infer whether eating chocolate drinking causes lowered risk of stroke. To do this, we compare a large group of chocolate eaters with another large group of non-chocolate eaters. Part A: Which is the control group? Which is the treatment group? Part B: Would it make sense to allow the study participants to choose which group they are in? Explain the possible consequences of such an approach. Part C: As researchers, it is our job to create control and treatment groups so that confounding variables, such as education or income, do not interfere with the outcome. What should be our goal as we decide how to assign subjects to groups? In controlled statistical experiments, researchers try to minimize the effects of confounding variables by creating treatment and control groups that are as similar as possible. Groups are similar when confounding variables are held constant (on average) from group to group. 5) You have gathered a group of volunteers who are all in their early 20s. They come from various socioeconomic backgrounds, and none eat chocolate, but they are all willing to try it for the sake of the experiment. You want to divide them into two groups (control and treatment), and you will monitor their health for four decades. Part A: Devise a method for dividing the volunteers into groups that will minimize confounding in the long-term results. Be specific. Part B: What are the explanatory and response variables in this study? Are they quantitative or categorical? Part C: With your method of dividing the volunteers into groups, suppose that the decades-long study concludes with a significantly lower rate of stroke among those who ate chocolate. Write a sentence that infers a conclusion for the study. The Charles A. Dana Center at The University of Texas at Austin 23 Version 1.0 Statistical Reasoning Student Pages 15.B, All things in moderation The Charles A. Dana Center at The University of Texas at Austin 24 Version 1.0 Statistical Reasoning Suggested Instructor Notes 15.B, All things in moderation Lesson 15, Part B All Things in Moderation Overview and student objectives Overview Lesson Length: 25 minutes In this lesson, students learn about confounding variables that can keep us from making causal inferences in statistical studies. They consider statistical studies that have documented associations between chocolate consumption and positive health benefits. Many studies conducted on this subject are observational and in such cases we are tempted to infer cause when cause has not been established. Students learn about controlled studies which attempt to hold confounding variables constant so that they can make causal conclusions. In later lessons, students will consider the placebo effect as another confounding variable in statistical studies and how to control this with placebos and blinding. Prior Lesson: Lesson 15, Part A, “The Video Game Diet” Next Lesson: Lesson 15, Part C, “The Power of the Pill” (25 minutes) Constructive Perseverance Level: 1 Theme: Health Literacy Goal: Evaluation Quantitative Reasoning Outcomes: S1, S2 Related Foundation Outcomes: C1, C2 Objectives Students will understand that: • Conclusions about cause and effect cannot be made in observational studies because of confounding variables. • Conclusions about cause and effect can be made in experimental studies where confounding is controlled. • Experimental groups created through random assignment can usually control variability among confounding variables. Students will begin to be able to: • Analyze a statistical study and identify possible confounding variables. • Decide when confounding variables restrict conclusions about cause and effect. • Design an experiment that allows conclusions about cause and effect. Suggested resources and preparation Materials and technology • Computer, projector, document camera • Preview Assignment 15.B • Student Pages for Lesson 15, Part B • Practice Assignment 15.B The Charles A. Dana Center at The University of Texas at Austin 25 Version 1.0 Statistical Reasoning Suggested Instructor Notes 15.B, All things in moderation Prerequisite assumptions Students should be able to: • Distinguish observational studies from experimental studies. • Identify explanatory and response variables. • Determine when variables are quantitative or categorical. • Define confounding. Making connections This lesson: • Connects back to Lessons 3 and 13 activities about samples, populations, and numerical summaries. • Connects forward to more on statistical studies (Lessons 19–21). Background context In Preview Assignment 15.B, students were asked to research information about confounding variables. Suggested instructional plan Frame the lesson (2 minutes) ThinkPairShare • Ask students to read the introductory information and respond to question 1. Take student responses. • Transition to the lesson activities by briefly discussing the Objectives for the lesson. Lesson activities (18 minutes) Group Work Guiding Questions Question 2 • Give students time to discuss this question and think about the problems associated with assigning a cause. This is the first time confounding has been directly addressed, and we want them to discover it on their own if they can. o “If two things happen together, does one always cause the other?” [Answer: No.] o “Can you think of other things that chocolate eaters might have in common?” [Answers may vary. Sample answer: They may have other dietary commonalities.] The Charles A. Dana Center at The University of Texas at Austin 26 Version 1.0 Statistical Reasoning Suggested Instructor Notes 15.B, All things in moderation Question 3 • Students are given their first concrete evidence of confounding variables (e.g., education) in this situation. Education can lead to many things, including a long list of other possible confounding variables. • Encourage students to think of other variables that are associated with being educated and also might contribute to better health. • The term confounding implies a sense of confusion in reasoning. The more confounding variables there are, the less certainty we have in drawing conclusions about cause. Question 4 • Here students begin to think through the process of designing their own study. This question leads them to think about the ideas of treatment and control groups. • Allow students to work together and struggle on these questions. As a group they need to come to the point where they understand that if people choose their own groups, the groups may end up being quite different, introducing confounding variables. o “If people choose their own groups, who will be likely to choose the chocolate eating group?” [Answer: Chocolate eaters.] Guiding Questions o “If chocolate eaters all go to the chocolate eating group, will they have things in common that could confound the study? What might they have in common?” [Answers will vary. Sample answer: Maybe gender.] o “What would help control confounding variables best: having very similar groups or very different groups?” [Answer: Groups should be similar to each other.] Question 5 • Students work together to devise a strategy for creating groups that are similar. Students tend to answer questions like this by saying something like “randomly divide participants into groups.” While this is correct, it is not specific as to how it is to be done. Encourage them to design a specific method to accomplish the grouping. Will they toss a coin? Maybe they will draw names out of a hat. There are many ideas that would work. • Remember that we want to make an inference about a treatment in this study, so a representative sample is less important than having groups that are similar. Still, representative samples are nice when we can have them, and larger samples are always a good thing. • Students sometimes have trouble deciding what the variables are in a study, especially when they are categorical. The explanatory variable, chocolate eating, is easy enough, but they may say that the response The Charles A. Dana Center at The University of Texas at Austin 27 Version 1.0 Statistical Reasoning Suggested Instructor Notes 15.B, All things in moderation variable is stroke risk. Remind them that the response variable is measured for each member in the sample. We do not measure rates from single individuals. • Guiding Questions In Part C, students practice making a statistical inference. Because confounding variables were controlled, it is appropriate to make a conclusion that indicates a causal relationship between chocolate eating and lower rates of stroke. Check with them to verify that their conclusion indicates cause. o “Is this an experimental or observational study?” [Answer: Experimental.] o “What type of statistical study allows us to make a cause and effect conclusion?” [Answer: Experimental.] Wrap-up/transition (5 minutes) Wrap-up • Take a few moments to go over the main ideas of this lesson and be sure the statistics dictionary is complete. o In experiments, we attempt to control confounding variables by holding them constant from group to group. This is done by making groups as similar as possible. o One way that researchers can make groups similar is to randomly assign subjects to groups. A method of assignment is random if each group has an equal chance of receiving any member. Transition • Ask students to reflect on Lesson 15, Parts A and B, and write their reflections on a 3x5 card. List one concept that is clear (or not clear) after the two lessons. • Have students refer back to the Objectives for the lesson and check the ones they recognize from the activity. Alternatively, checking objectives may be done throughout the lesson. Suggested assessment, assignments, and reflections • Give Practice Assignment 15.B. • Give the Preview Assignments, if any, for the lesson activities you plan to complete in the next class meeting. The Charles A. Dana Center at The University of Texas at Austin 28 Version 1.0 Statistical Reasoning Suggested Instructor Notes 15.B, All things in moderation Lesson 15, Part B All Things in Moderation – ANSWERS Chocolate has been a favorite food of humans for over 3,000 years, originating from Central America and Mexico, where it was enjoyed as a drink. It was thought to have certain medicinal properties and was even considered valuable enough to be used as money.1 Lately, studies have emerged that also promote the health benefits of chocolate. 1) Have you heard of any health benefits from eating chocolate? If so, what were they? Answers will vary. Sample answer: Chocolate is an antioxidant; lower cholesterol; reduce cognitive decline as you age; reduce risk of cardiovascular problems, heart disease, stroke. Credit: ThinkStock Objectives for the lesson You will understand that: o Conclusions about cause and effect cannot be made in observational studies because of confounding variables. o Conclusions about cause and effect can be made in experimental studies where confounding is controlled. o Experimental groups created through random assignment can usually control variability among confounding variables. You will begin to be able to: o Analyze a statistical study and identify possible confounding variables. o Decide when confounding variables restrict conclusions about cause and effect. o Design an experiment that allows conclusions about cause and effect. In this lesson, we consider how conclusions in statistical studies can be uncertain when researchers do not know why a particular outcome has happened. 2) A research study in Sweden asked the research question: “Does chocolate reduce the risk of stroke?”2 1 Benson, A. (2008, March 1.) A brief history of chocolate. Retrieved November 24, 2014, from http://www.smithsonianmag.com/arts-culture/a-brief-history-of-chocolate-21860917/?no-ist. 2 Larsson, S. C., Virtamo, J., & Wolk, A. (2012). Chocolate consumption and risk of stroke: A prospective cohort of men and meta-analysis. Neurology, 79(12), 1223-–1229. The Charles A. Dana Center at The University of Texas at Austin 29 Version 1.0 Statistical Reasoning Suggested Instructor Notes 15.B, All things in moderation A sample of 37,000 Swedish men were divided into four groups based on their normal chocolate consumption, ranging from those who ate no chocolate to the group who ate the most (63 grams per week). It is important that the study included so many people because it makes the results more reliable. Participants were followed for 10 years. The study concluded that men who ate more chocolate had a lower risk of experiencing a stroke. Part A: Did this study prove that eating chocolate causes a lower risk of stroke? Explain your response. Answer: No. Sample explanation: This is an observational study so we cannot infer causation. There may be other variables that can explain lowered risk of stroke among chocolate eaters. Part B: Offer an alternate explanation for this lowered risk among chocolate eaters. Answers will vary. Sample answer: Maybe people who ate more chocolate also tended to eat something else that could be contributing to lower risk of stroke, or maybe they tended to not eat something that may contribute to a higher risk of stroke. 3) In Preview Assignment 15.B, you read about confounding variables, or factors that give alternate explanations for an effect. When there are confounding variables, researchers cannot make inferences about cause and effect. Another article3 pointed out some important facts in the study described above. The men who ate the most chocolate tended to be younger, were more educated, and were less likely to smoke or high blood pressure. They also reported eating more vegetables and drinking more wine. All of these factors have positive health benefits. Part A: Explain how this information sheds light on the study that says that eating chocolate lowers the risk for stroke? Does it make us more or less certain about the effects of eating chocolate? Answers will vary. 3 Goodman, B. (2013, January 7). Don’t fudge the facts on chocolate studies. Retrieved November 24, 2014, from http://healthjournalism.org/blog/2013/01/dont-fudge-the-facts-on-chocolate-studies. The Charles A. Dana Center at The University of Texas at Austin 30 Version 1.0 Statistical Reasoning Suggested Instructor Notes 15.B, All things in moderation Sample answer: It tells us that chocolate eaters have other attributes, which may also contribute to lower stroke risk. These other factors make us less certain about the effects of chocolate eating. Part B: Chocolate eaters tended to be more highly educated. What other confounding variables arise from this fact that may also cause a lowered risk of stroke? (What other things do people who are better educated tend to have in common?) Answers will vary. Sample answer: People who are more educated often make more money and thus have better health care. More educated people may eat better and exercise more. Education can lead to many behaviors that are beneficial for good health. All of these are confounding variables. Cause and effect conclusions cannot be made in observational studies because of confounding variables. Experiments allow such conclusions by holding confounding variables constant among several groups. To do this, researchers make sure that groups are similar in terms of education, age, eating habits, and so on. When groups are similar, confounding variables are controlled (not varying), and it becomes easier to argue that a treatment (like chocolate) is the cause of an observed effect (like fewer people having a stroke). To control for the confounding variables, researchers divide participants into similar groups. Some groups are treatment groups—they receive a real treatment. Other groups are control groups—they receive a placebo (a fake treatment) or nothing at all. Control groups are used as a basis for comparison to the treatment groups. 4) We want to design a new study to infer whether eating chocolate drinking causes lowered risk of stroke. To do this, we compare a large group of chocolate eaters with another large group of non-chocolate eaters. Part A: Which is the control group? Which is the treatment group? Answer: Control group: Non-chocolate eaters. Treatment group: Chocolate eaters. Part B: Would it make sense to allow the study participants to choose which group they are in? Explain the possible consequences of such an approach. Answer: No. The Charles A. Dana Center at The University of Texas at Austin 31 Version 1.0 Statistical Reasoning Suggested Instructor Notes 15.B, All things in moderation Sample explanation: Allowing participants to choose their group would be a mistake. Chocolate eaters would choose the chocolate-eating (treatment) group, and nonchocolate eaters would pick the other (control) group. Chocolate eaters probably have other factors in common that could make the causes of any observed results unclear. Part C: As researchers, it is our job to create control and treatment groups so that confounding variables, such as education or income, do not interfere with the outcome. What should be our goal as we decide how to assign subjects to groups? Answer: Our goal should be to make the groups similar so that the confounding variables are held constant from group to group. In controlled statistical experiments, researchers try to minimize the effects of confounding variables by creating treatment and control groups that are as similar as possible. Groups are similar when confounding variables are held constant (on average) from group to group. 5) You have gathered a group of volunteers who are all in their early 20s. They come from various socioeconomic backgrounds, and none eat chocolate, but they are all willing to try it for the sake of the experiment. You want to divide them into two groups (control and treatment), and you will monitor their health for four decades. Part A: Devise a method for dividing the volunteers into groups that will minimize confounding in the long-term results. Be specific. Answers may vary. Sample answer: Use a random number generator to separate them into two groups. Volunteers who get an odd number will be in the control (no-chocolate) group; those with an even number will be in the treatment group. Sample answer: Carefully design the groups to be similar by matching them regarding diet, exercise habits, etc., Then randomly assign one person of each similar pair to the treatment group and the other to the control group. Part B: What are the explanatory and response variables in this study? Are they quantitative or categorical? Answer: The explanatory variable is eating chocolate (yes or no). The response variable is stroke (yes or no). Both variables are categorical. The Charles A. Dana Center at The University of Texas at Austin 32 Version 1.0 Statistical Reasoning Suggested Instructor Notes 15.B, All things in moderation Part C: With your method of dividing the volunteers into groups, suppose that the decades-long study concludes with a significantly lower rate of stroke among those who ate chocolate. Write a sentence that infers a conclusion for the study. Answer: With this carefully designed experimental study, we can state that there is evidence that eating chocolate is the cause of lower risk for stroke. The Charles A. Dana Center at The University of Texas at Austin 33 Version 1.0 Statistical Reasoning Suggested Instructor Notes 15.B, All things in moderation The Charles A. Dana Center at The University of Texas at Austin 34 Version 1.0 Quantitative Reasoning Practice Assignment 15.B Practice Assignment 15.B 1) Researchers conducted a study to learn how early childhood education relates to a child’s future earning potential. They gathered a random sample of 200 5-year-old children and divided them into two groups. Children who had preschool education were gathered into the first group. All other children were placed in the second group. By age 30, the subjects who received preschool education were 80% less likely to have received welfare in their 20s. Part A: This study was ___________. a) Observational b) Experimental Part B: Having a preschool education (yes or no) is one variable in this study. Which is it? a) Response b) Explanatory Part C: Receiving welfare in one’s 20s (yes or no) is another variable in this study. Which is it? a) Response b) Explanatory Part D: List possible confounding variables. Part E: Can researchers conclude from this study that preschool education causes people to be less likely to receive welfare in their 20s? a) Yes b) No The Charles A. Dana Center at The University of Texas at Austin 35 Version 1.0 Quantitative Reasoning Practice Assignment 15.B Part F: Explain your answer to Part E. 2) The damaging effects of drinking too much alcohol have been observed for centuries. Only lately have statistical studies shown that drinking alcohol in moderate amounts is associated with positive health benefits.1 These studies do not always explain why these benefits occur. In one study focusing on the effects of drinking wine, a representative sample of adults was divided into three groups: non-wine drinkers, moderate wine drinkers, and heavy wine drinkers. Participants were tracked for decades recording wine consumption and heart disease. The moderate wine drinkers had the lowest risk of heart disease. Part A: Did this study prove that moderate wine drinking causes a lower risk of heart disease? a) Yes b) No Part B: Offer an alternate explanation for this lowered risk among moderate wine drinkers. 1 Mayo Clinic. (2014, February 11.) Alcohol use: If you drink, keep it moderate. Retrieved November 23, 2014, from http://www.mayoclinic.org/healthy-living/nutrition-and-healthy-eating/in-depth/alcohol/art20044551. The Charles A. Dana Center at The University of Texas at Austin 36 Version 1.0 Quantitative Reasoning Practice Assignment 15.B Part C: In 2001, Danish researchers observed that moderate wine drinkers are better educated, on average.2 Explain how this information sheds light on the study that says that moderate wine drinkers have lower risk for heart disease. Does it make us more or less certain about the effects of moderate wine drinking? a) This observation makes us more certain that drinking a moderate amount of wine lowers the risk for heart disease. b) This observation makes us less certain that drinking a moderate amount of wine lowers the risk for heart disease. Part D: Besides education, think of other confounding variables that might cause a lowered risk of heart disease among moderate wine drinkers. 3) We want to design an experimental study to infer whether moderate wine drinking causes lowered risk of heart disease. To do this, we plan to compare participants who drink a moderate amount of wine with participants who do not drink wine. Part A: Which is the control group? Which is the treatment group? a) Non-wine drinkers are the control group, and moderate drinkers are the treatment group. b) Moderate wine drinkers are the control group, and non-wine drinkers are the treatment group. Part B: Would it make sense to allow the study participants to choose which group they are in? a) Yes b) No 2 Mortensen, E. L., Jensen, H. H., Sanders, S. A., & Reinisch, J. M. (2001). Better psychological functioning and higher social status may largely explain the apparent health benefits of wine. Archives of Internal Medicine, 161(15), 1844-1848. The Charles A. Dana Center at The University of Texas at Austin 37 Version 1.0 Quantitative Reasoning Practice Assignment 15.B Part C: As researchers, it is our job to create control and treatment groups so that confounding variables, such as education or income, do not interfere with the outcome. What should be our goal as we decide how to assign subjects to groups? a) To make sure that everyone in the control group is similar to one another and that everyone in the treatment group is similar to one another. b) To make sure that the composition of the control group is similar to the composition of the treatment group. Part D: You have gathered a group of volunteers that are all in their early 20s. They come from various socioeconomic backgrounds, and none drink wine, but they are all willing to try it for the sake of the experiment. You want to divide them into two groups (control and treatment), and you will monitor their health for four decades. Which of the following is an acceptable method for dividing the volunteers into groups that will minimize confounding in the long-term results. a) Put all of the men into the treatment group and all of the women into the control group. b) Put all of the women into the treatment group and all of the men into the control group. c) Put all of the participants under the age of 40 in the treatment group and all of the participants over the age of 40 in the control group. d) Put all of the participants under the age of 40 in the control group and all of the participants over the age of 40 in the treatment group. e) Randomly assign each participant to either the treatment group or the control group. Part E: With your method of dividing the volunteers into groups, suppose that the decades-long study concludes with a significantly lower rate of heart disease among those who drank moderate amounts of wine. Write a sentence that infers a conclusion for the study. The Charles A. Dana Center at The University of Texas at Austin 38 Version 1.0