Introduction

1. "There are three types of lies: lies, damned lies and statistics" Mark Twain
    
   1. 400,000 people die per year from smoking related causes. (Source)
       
      1. To put it in perspective there are about 250 million Americans and in 1997 there were 2.3 million deaths. So, about 17% of all deaths are smoking related.
      2. Who is a smoker?
      3. How do you know what someone dies of?
      4. Smokers tend to be poorer, less educated and drink more alcohol than non-smokers. Does this complicate estimating the the number of smoking related deaths?
      5. What does smoking related mean?
      6. Does this mean that if everyone stopped smoking today that annual mortality would decrease by 400,000?
      7. What does it mean?
          
   2. A black person is nearly 4 times more likely to be executed under death penalty laws than a white person. (How do you figure?)
       
      1. But I thought you just said the majority of people executed were white?
      2. Why is this so? Does this statistic tell you? (12% of Americans are black but 36% of executed people are black.)
      3. If you think that there is a racial bias in death penalty cases which statistic would you prefer to cite? What if you didn't think there is a bias?
          
   3. Motorcycle Helmets (read The Case of No Helmets (the article discussed on page 13 of the text) and the CDC Head Injury report)
       
      1. Dick Teresi states "Some studies indicate that there are fewer motorcycle deaths per one-million residents in states with helmet laws than in states where helmets are not mandatory." He says this is a CDC statistic from the years 1979-1986.
      2. Mr. Teresi thinks this rate should be expressed in number of motorcycle deaths per 10,000 registered motorcycles, not per one-million residents. Using this criteria he finds that death rates in helmet laws states is lower than in non-helmet law states (3.05 versus 3.38).
      3. Which statistic would you use if you were in favor of motorcycle helmet laws? What if you were against them?
          
2. Simple statistics
    
   1. Some statistics are known without error
       
      1. Dow Jones index
      2. IQ
      3. 611 people have been executed via the death penalty since 1976 (as of 1/11/2000, Source)
      4. The majority (55%) of those executed under death penalty laws are white
          
   2. Some statistics are estimates
       
      1. An aspirin a day decreases your risk of heart attack 44% (Source)
      2. unemployment rates
      3. polling results
          
3. Descriptive Statistics
    
   1. simple statistics, summaries
   2. graphs
         
4. Inferential Statistics
    
   1. uses data collected from samples to make generalizations about populations
   2. subjects can be anything: people, manufactured parts, car accidents, etc. (see the first paragraph of Section 1.3 of the text on page 4)
   3. A dataset: how data is recorded on paper or in a computer database. The most common type a dataset is a table with 1 row per subject and one column per variable. The data itself resides within the table.
   4. Example: Say we are interested in the average income of 1990 WSU graduates
       
      1. Population: all 1990 graduates
      2. Sample: those graduates we are able to contact
      3. Variable: yearly income in dollars
      4. Data: the set of all incomes reported by the sample
          
   5. Example: We want to estimate the effect of regular aspirin use on heart attack risk (Physicians' Health Study)
       
      1. Population: all Americans
      2. Sample: 22,071 male physicians
      3. Variables:
      4. Data:

5. Types of Inferential Statistics
    
   1. estimation: estimate of a parameter of a population
      1. Mean income of a population
      2. Percent of people that will vote for a certain candidate.
      3. How much will a person's cholesterol level drop if they take the drug Lovastatin?
      4. How many people die per year from smoking?
          
   2. decisions
      1. Does regular use of aspirin decrease the risk of heart attacks?
      2. Is the death penalty racially biased?
      3. Does prayer increase survival rates?
      4. Should mammograms be performed annually on all women over 45
          
   3. prediction
        
6. Reliability
    
   1. Much of statistics focuses on measuring the reliability of inferences
   2. How reliable are the results of a census?
   3. Polling data: usually state that poll results are +/- 3%  

E-mail Mr. Callahan at stat110@edcallahan.com with questions or comments about this web site or about the class itself.