Understanding Research Methods: Observational Studies and Experiments - Prof. Murali S. Sh, Study notes of Business Statistics

Download Understanding Research Methods: Observational Studies and Experiments - Prof. Murali S. Sh and more Study notes Business Statistics in PDF only on Docsity! Chapter Goals To understand the importance of looking at studies critically. Formulate Theories Collect Data Summarize Results Interpret Results/Make Decisions Chapter 3: Observational Studies and Experiments 3–1 The Language of Studies Observation versus Experimentation Definition In an experiment, the researcher actively imposes some treatment on the units or subjects in order to observe the response. Definition In an observational study, the researcher simply observes the subjects or units and records the variables of interest. Note: Are there circumstances when an experiment is not feasible? Chapter 3: Observational Studies and Experiments 3–2 Relating Variables Definition A response variable measures an outcome of the study. It is sometimes referred to as the dependent variable. Definition An explanatory variable or factor is a variable that is thought to explain or cause the observed outcome in the response variable. The explanatory variables are sometime referred to as independent variables. Definition The possible values of the explanatory variables are called levels, and a specific combination of the levels of an explanatory variable is called a treatment. Example Chapter 3: Observational Studies and Experiments 3–3 Let’s do it! 3.2 Research done at Harvard showed that regular soda drinkers may have a tendency toward the weak and brittle bones associated with osteoporosis. The researchers surveyed 2622 women who were active athletes in college and classified them as to whether or not they regularly drank soda. The proportion of women in each group who suffered from bone fractures was determined. It was found that those who drank soda regularly were twice as likely to suffer from bone fractures than those who did not drink regularly. 1. Is this an observational study or an experiment? 2. What is the population under study? 3. What was the explanatory variable? 4. What was the response variable? 5. Suppose the 2622 women were selected as follows: From a list of all women sports events available at Harvard, three events were selected at random. Then all women athletes who participated in these three events were contacted for the study. What type of sampling technique is illustrated here? Chapter 3: Observational Studies and Experiments 3–4 Let’s do it! 3.5 An experiment has been conducted to study the effect of temperature and type of oven on the lifetime of a metal clutch component. Four types of ovens and three temperature levels were used in the experiment. Two clutch components were assigned randomly to each combination of type of oven and temperature level. 1. For this experiment: a. what are the experimental units? b. what is the response variable? c. what are the factors and the levels of the factors? d. list the treatments. 2. Draw the design layout table for this situation. Chapter 3: Observational Studies and Experiments 3–9 Planning Experiments Exercise A researcher believes that taking a daily dose of vitamin C will help reduce the incidence of colds in elementary school children. To assess this claim, 100 elementary school children, who volunteered and for whom parental consent was received, were given a daily dose of vitamin C for a three-month winter period. The number of colds during this season was recorded for each subject. For this experiment we have 100 subjects, one response variable, the number of colds, and one explanatory variable, vitamin C, at one level, a daily dose. Results # of Colds 0 1 2 3  4 Frequency 45 28 12 7 8 Based on this data, does taking a daily dose of Vitamin C help? Chapter 3: Observational Studies and Experiments 3–10 Definition A treatment group is a group of subjects or experimental units that receive an actual treatment. Definition A control group is a group of subjects or experimental units that are treated identically in every way, except that they do not receive an actual treatment. Exercise Can you now think of a better way of assigning the 100 subjects? How would you decide which subject gets what treatment? Chapter 3: Observational Studies and Experiments 3–11 Definition In a completely randomized experiment, experimental units are randomly assigned to the treatments. Exercise A researcher has a treatment group and a control group. The control group will receive a placebo. The subjects will be randomly assigned to one of the two groups. Those in the treatment groups receive a bottle of medication marked DRUG, while those in the control group receive a bottle of medication marked PLACEBOS. Comment on this experimental design? Chapter 3: Observational Studies and Experiments 3–12 Definition The placebo effect is a phenomenon in which even administration of an inert drug improves the condition of the subjects. Definition Experimenter bias is the distortion that can arise on the part of the experimenter due to how the subjects are assigned to the groups, which variables are measured and how they are measured, and how the results are interpreted. Definition A single-blind experiment is one in which the subjects are ignorant of which treatment they receive. Definition A double-blind experiment is one in which neither the subjects nor those working with the subjects knows who is receiving which treatment. Definition If at least two units are assigned to each treatment combination, we have replication in an experiment. Definition An experimental design is balanced when each treatment combination has the same number of experimental units. Chapter 3: Observational Studies and Experiments 3–13 Chapter Summary Basic Principles for Design of Experiments: Randomization Use randomization to assign units to the groups. Randomization tends to produce groups of experimental units that are similar with respect to potential confounding factors. Randomization ensures that the experiment does not intentionally favor one treatment over another. Control Control the effects of confounding variables by using a comparative design. If all subjects are treated exactly the same across the groups, except for the actual treatment they received, potential confounding variables should affect both groups equally and tend to cancel each other out when comparing the results from two groups. Blinding If possible, give a placebo to the control group. Neither the subjects nor anyone working with the subjects should know who is receiving the treatment and who is getting the placebo. Replication If possible, assign at least two units to each treatment combination to help assess the natural variation in the responses Chapter 3: Observational Studies and Experiments 3–14