Example – Genetic CounselingExample 1 – PrognosisExample 2 – Federal Drug TestingExample 3 – Smoking and CancerExample 4 – Justice versus MedicineExample 5 – Investigation of the Portacaval ShuntReported Enthusiasm for ShuntDesignNo controlsA Feel for Things One goal is to inform decision making, as in the examples described in previously:- Description – we use the values of statistics from a sample to make estimates about unknown population parameter values. - Inference making – through the fitting and comparison of competing models of the data, we obtain a comparison (hypothesis test) of competing explanations (hypotheses) of the phenomena we have observed. Example 7 -ExampleExample, continued -A simple random sample of n2 =25 students is taken from the N2 =220 students at a second, independent, school. These latter 25 were given the blood test as above. Using techniques of statistical hypothesis testing, a conclusion is drawn regarding the similarity of the blood levels at the two schools.Key PointsPubHtlth 540 – Fall 2010 Course Introduction Page 1 of 21 Course Introduction “Very true,” said the Duchess: “flamingoes and mustard both bite. And the moral of that is – “’Birds of a feather flock together. ‘ “ “Only mustard isn’t a bird,” Alice remarked. “Right, as usual,” said the Duchess: “what a clever way you have of putting things!” - Alice in Wonderland The course introduction outlines the direction of the entire course, using a “course roadmap.” Statistical literacy is introduced using several examples. You might discover that we are often poor at evaluating probability! Last, a brief overview of each unit is provided. Nature Population/ Sample Observation/ Data Relationships/ Modeling Analysis/ SynthesisPubHtlth 540 – Fall 2010 Course Introduction Page 2 of 21 Table of Contents Topic 1. Course Roadmap ……………………………………………………. 2. A Feel for Things …………………………………………….……. 3. Overview, Unit by Unit ………………………………………...…… 3 5 12 Key Points ……………………………………………………………... 21 Nature Population/ Sample Observation/ Data Relationships/ Modeling Analysis/ SynthesisPubHtlth 540 – Fall 2010 Course Introduction Page 3 of 21 1. Course Roadmap Nature Nature is full of variation. Variation might be from time to time, from person to person, or from one repeated measurement to the next. Variation might be from one treatment to the next or from one exposure to the next. Which variation is “real” and which variation is “natural”? Do we even know what we’re talking about when we distinguish “real” from “natural”? Populations/ Sample A population is a class of individuals. An example is the collection of individuals who voted in the 2008 U.S. presidential election. Numerical facts about a population are called parameters. If we could study a population by examining each and every member, we would be doing a census. This course is not about censuses. More often, what we can examine is only a part of a population; that part is called a sample. Numerical facts about a sample are called statistics. Statistics from a sample are used to make generalizations to the population. This is called inference. Observation/ Data Observation and data may not be the same. What does your mind’s eye “register” when you observe a flower? You might describe the flower as red, with 5 petals, and having a strong aroma. “Red”, “5 petals”, “strong aroma” are your data. Data are the result of selection (which attributes of the flower matter to you in the first place?) and measurement (what value scheme are you using?). Just think of the many attributes of the flower that were not selected as your data! A variable is something whose value can vary. Data are the values you obtain by measurement of the variable. “Color” is a variable. “Red” is a data value. Nature Population/ Sample Observation/ Data Relationships/ Modeling Analysis/ SynthesisPubHtlth 540 – Fall 2010 Course Introduction Page 4 of 21 Relationships/ Model A relationship exists between two variables if they covary (eg; – the relationship between excessive sun exposure and occurrence of skin cancer) Statistical modeling is used to discover relationships. Beginning with the data, models are fit to the data and not the other way around! In fact, there might well be several models that are a good description of the available data. A good model is one that (i) explains a good amount of the variability in the data (adequacy); and is then (ii) minimally adequate (parsimony), meaning: it represents your best understanding of the factors that are related to your response variable as simply as possible. Analysis/ Synthesis The existence of a relationship does not mean there is causality. Nature Population/ Sample Observation/ Data Relationships/ Modeling Analysis/ SynthesisPubHtlth 540 – Fall 2010 Course Introduction Page 5 of 21 2. A Feel for Things A variety of illustrations provide a feel for things. Example – Genetic Counseling A couple has a baby with a genetic defect. They are considering having another baby. What is the likelihood that the second child will have a genetic defect also? Example 1 – Prognosis A physician is considering several therapies for the treatment of a patient. Which therapy should be used? Each therapy produces a result that is somewhere between success and failure. The final choice is “weighed” against the others. Probabilities are a tool in decision making. Example 2 – Federal Drug Testing Is a food additive carcinogenic? An investigator explores this in an experiment that compares two groups. Only some of the controls develop cancer. Only some of the treated individuals develop cancer. Is the excess number of cancers among treated individuals meaningful? Nature Population/ Sample Observation/ Data Relationships/ Modeling Analysis/ SynthesisPubHtlth 540 – Fall 2010 Course Introduction Page