BMS 511 Biostats & Statistical Analysis PDF

BMS 511 Biostats & Statistical Analysis Chapter 7 Designing experiments Guang Xu, PhD, MPH Assistant Professor of Biostatistics and Public Health College of Osteopathic Medicine Marian University...

BMS 511 Biostats & Statistical Analysis Chapter 7 Designing experiments Guang Xu, PhD, MPH Assistant Professor of Biostatistics and Public Health College of Osteopathic Medicine Marian University Copyright © 2018 W. H. Freeman and Company  Previous Learning Objectives Determine samples and observational studies Observational study versus experiment Population versus sample Randomness and bias The simple random sample (SRS) Other probability samples Sample surveys Comparative observational studies Copyright © 2018 W. H. Freeman and Company  Learning Objectives Designing experiments Terminology of experiments Randomized, comparative experiments Completely randomized designs Block designs Matched pairs designs Double-blind experiments Ethics and experimentation Copyright © 2018 W. H. Freeman and Company  Terminology The individuals in an experiment are the experimental units. If they are human, we call them subjects. The explanatory variables in an experiment are often called factors. A treatment is any specific experimental condition applied to the subjects. If an experiment has several factors, a treatment is a combination of specific levels of each factor. – The factor may be the administration of a drug. – One group of people may be placed on a diet/exercise program for 6 months (treatment), and their blood pressure (response variable) would be compared with that of people who did not diet or exercise. Copyright © 2018 W. H. Freeman and Company  Randomized, comparative experiments Experiments compare the response to a given treatment versus: another treatment the absence of treatment (often called a control) a placebo (a fake treatment) Experiments randomize the assignment of subjects to treatments. Experiments use replication: several or many individuals are studied. Copyright © 2018 W. H. Freeman and Company  Inventing experimental design (1 of 2) Ronald Fisher was sent to a UK agricultural station to evaluate the effects of fertilizers. He found decades worth of bad data. – Fertilizer had been applied to a field one year and not in another to compare the yield of grain produced in the two years. – Fertilizer was applied to one field and not to a nearby field in the same year. Copyright © 2018 W. H. Freeman and Company  Inventing experimental design (2 of 2) Satellite and soil recordings show that soil types and moisture vary, F F F F F F even within a single field. F F F F F F F F  Solution: Randomized F F F F F comparative experiments. F F F F F F F F He selected many fields and F F F F F randomly assigned the fields to F F F F receive fertilizer or not. Grain yield was then compared for the two conditions. Copyright © 2018 W. H. Freeman and Company  Importance of design (1 of 2) Gastric freezing was once a recommended treatment for peptic ulcers. Patients would swallow a balloon through which a refrigerated liquid was pumped for an hour to cool the stomach. The treatment was shown to be safe and significantly reduce ulcer pain, and it was widely used for years (2,500 gastric freeze machines sold and 15,000 patients were chilled). Copyright © 2018 W. H. Freeman and Company  Importance of design (2 of 2) A randomized comparative experiment was later performed to compare the effect of gastric freezing with that of a placebo: – 28 of the 82 patients subjected to gastric freezing improved – 30 of the 78 patients in the control group improved Because the rates were so similar, gastric freezing was then abandoned. Copyright © 2018 W. H. Freeman and Company  Completely randomized designs In a completely randomized experimental design individuals are randomly assigned to groups, then the groups are assigned to treatments, completely at random. Copyright © 2018 W. H. Freeman and Company  Block designs (1 of 2) In a block design, subjects are divided into blocks (groups sharing a given characteristic) before the randomization, in order to account for possible differences between the blocks. The explanatory variables in an experiment are often called factors. The blocks are NOT random. Randomization occurs inside each block. A block design lets us choose how many individuals of each block will receive each treatment. Copyright © 2018 W. H. Freeman and Company  Block designs (2 of 2) Main Outcome Measure: Mean number of symptom- free days per 2 weeks during the peak winter season (November-February), assessed by blinded interviews. Copyright © 2018 W. H. Freeman and Company  Matched pairs designs Choose pairs of subjects that are closely matched (such as twins). Within each pair, randomly assign who will receive which treatment. Or give the two (or more) treatments to each subject over time, in random order, so we have repeated measures for each subject. Each subject is given two chili bowls: version A and version B. Subjects eat and rate both versions. Half of all subjects are given version A first then version B. The other half are given version B first then version A. Copyright © 2018 W. H. Freeman and Company  The placebo effect (1 of 2) Improvement in health or perceived condition may be due not to any active treatment, but only to the patient’s belief that he or she is being cared for or helped. therapeutic results on up to 35% of patients famous placebo: kiss/blow/hug/Band-aid to help kids cope with minor injuries neural response to the placebo effect seen as early as the spinal cord “negative placebo effect” also observed; labeling an active drug a placebo lowers its effectiveness Copyright © 2018 W. H. Freeman and Company  The placebo effect (2 of 2) Science (2009), DOI: 10.1126/science.1180142 Sci Transl Med (2014), DOI: 10.1126/scitranslmed.3006175 Copyright © 2018 W. H. Freeman and Company  Double-blind experiments Bias, conscious or unconscious, from the placebo effect (subjects) or the experimenter is a challenge. “Blinding” can help against bias. A double-blind experiment is one in which neither the subjects nor the experimenter(s) know which individuals received which treatment until the experiment is completed. – However, subjects must be informed that they will get one of a number of treatments, and must consent to that condition (it would be unethical otherwise). Copyright © 2018 W. H. Freeman and Company  Examples of blinded studies (1 of 2) The Clinical Efficacy of Kinesio Tape for Shoulder Pain: A Randomized, Double-Blinded, Clinical Trial doi:10.2519/jospt.2008.2791 Efficacy of acupuncture for migraine prophylaxis: A single- blinded, double-dummy, randomized controlled trial doi: 10.1016/j.pain.2011.04.006 Copyright © 2018 W. H. Freeman and Company  Examples of blinded studies (2 of 2) FIGURE 2. Therapeutic Kinesio FIGURE 3. Sham Kinesio Tape Tape application application Copyright © 2018 W. H. Freeman and Company  Experimental design examples Six male runners participated in two running trials separated by one week. In random order, they consumed either chocolate milk or a Gatorade-like sports drink after their run. Protein synthesis was assessed three hours later. In a study of sickle-cell anemia, 152 patients were given the drug hydroxyurea, and 147 were given a placebo (dummy pill), based on random assignment. The researchers counted the episodes of pain in each subject at the end of the study. Copyright © 2018 W. H. Freeman and Company  Lack of realism (1 of 2) Is the sample representative of the target population? Random sampling is meant to gain information about a larger population. Is the treatment appropriate for the response you want to study? Copyright © 2018 W. H. Freeman and Company  Lack of realism (2 of 2) FDA requiring lower recommended dose for certain sleep drugs containing zolpidem (Jan. 10, 2013): “Since women eliminate zolpidem from their bodies more slowly than men, the FDA has notified the manufacturers that the recommended dose should be lowered for women.” Carcinogenicity studies administer high doses of a potential carcinogen to lab rats. Results don’t always apply to humans (e.g., saccharin delisted in 2000). Copyright © 2018 W. H. Freeman and Company  Ethics and experimentation Biology deals with life. Experimentations have an impact on live subjects and ecosystems. What rights do human subjects, animals, and ecosystems have? There is a difference between what can physically be done and what can be done ethically. When is it okay/not okay to include a placebo group? When should an experiment be interrupted? Personal standards vary, and extreme experimentations have occurred. Committees have been established to review all research proposals. Subjects must give “informed” consent. Copyright © 2018 W. H. Freeman and Company Short Break! SPSS SPSS SPSS  Exercise 1 For the following datasets, check to see if the data look normally distributed. Provide graphical evidence to support your claim. If the data do not look normally distributed, can anything be done to transform them? Provide graphical evidence to support your claims. You should be able to copy/paste into the software of your choice – it contains only one group of data Exercise Data 105.41 14.85 29.95 24.65 2.6 16.61 274.21 36.86 42.97 7.11 15.43 213.66 0.46 39.08 135.89 51.77 113.24 28.98 9.11 62.41 9.63 181.01 424.93 0.46 56.2 0.92 16.02 56.3 15.21 68.34 166.03 15.21 6.33 30.74 43.89 7.44 5.22 173.56 45.26 127.01 8.55 146.57 79.03 7.11 6.57 24.28 22.51 51.43 0.46 253.52 0.46 0.46 3.41 215.37 31.36 15.59 0.46 163.48 1.77 0.46 275.92 0.46 34.08 11.14 18.97 5.31 1.08 90.78 4.67 2.3 3.51 31.36 149.83 18.97 1.08 0.46 7.34 57.59 62.25 4.11 142.99 3.51 30.74 40.69  Group Project Exercise A chance to “apply” the statistics we have learned A chance to try to think critically oWhat’s the story? oWhat are the methods? oWhat are the results? oWhat are the conclusions? oWhat’s the future? oWhat do you think?  Learning Objectives Designing experiments Terminology of experiments Randomized, comparative experiments Completely randomized designs Block designs Matched pairs designs Double-blind experiments Ethics and experimentation Copyright © 2018 W. H. Freeman and Company