AMHERST COLLEGE BIOLOGY 28 - BIOSTATISTICS SPRING 2000

Instructor: Stephen George Office/Lab: Life Sciences 228/215 Office hours: Tuesdays 9:00 - 12:00, Fridays 1 - 3:30 Phone: extension 2477; E-mail; Home page

This course has additional materials on a "CourseInfo" site: http://courseinfo.amherst.edu/courses/BIOL-28-00S/

CONTENTS:

Course Outline

 I. PROBABILITY January 24 Introduction: variability in the life sciences Chapter 1 26 Combinations and probabilities Handouts 28 Conditional probabilities; Bayesian methods Chapter 14 II. Distributions and Confidence Intervals 31 Binomial distribution and confidence intervals Chapters 2, 3 February 2 Using Excel (Webster 102) 3 Problem set 1 due 4 Quiz 1 7 Normal distribution, confidence interval of the mean, t Chapters 4, 5 9 Presentations 11 Survival curves; introduction to paired comparisons Chapters 6, 7 14, 16 The Poisson distribution Chapter 28, handouts 17 Problem set 2 due 18 Quiz 2 III. Hypothesis testing 21 Hypothesis testing Chapters 10, 11 23 Presentations 25 Interpretation of p-values; multiple comparisons Chapters 12, 13 28, March 1 Comparisons of central tendency: parametric & non-parametric Chapters 23, 24, 25 2 Problem set 3 due 3 Quiz 3 IV. Clinical trials; linkage studies 6 Clinical trials; odds ratio; case control studies Chapters 8, 9 8 Presentations 10 More Bayesian methods: linkage, genetics Chapters 15, 16 V. Correlation and Regression 20, 22 Correlation Chapter 17 23 Problem set 4 due 24 Quiz 4 27 Regression Chapters 18, 19 29 Presentations 31 Determining sample size; statistical power Chapters 20, 21, 22 VI. Chi-square Methods April 3, 5 Chi-square; contingency tables Chapters 26, 27 6 Problem set 5 due 7 Quiz 5 VII Analysis of variance 10 ANOVA I Chapter 30 12 Presentations 14 ANOVA II Handouts 17, 19 Data transformations Handouts 20 Problem set 6 due VIII. Statistical modeling 21 Modeling I: Signals and noise Handouts 24 Modeling II: Neuroscience 26 Presentations 28 Quiz 6 May 1 Presentations 3 Presentations 5 Problem solving/ review for final exam

Resources

INTERNET RESOURCES FOR BIOLOGY 28 (BIOSTATISTICS), Spring 1997

Journal of Statistics Education, and on-line journal with access to datasets, including data on colleges from the US News & World Report's College issue and many other resources.
CHANCE News, a superb resource containing reviews of articles and data relevant to probability and statistics. This site contains 4 years of back issues of Chance magazine.
StatLib from Carnegie Mellon U., with info., links, and datasets.
UCLA Statistics Homepage, with useful case studies containing interesting data sets.
WWW Virtual Library: Biostatistics, from the University of Washington.
"Our Genes, Ourselves?", article in Biosciences by Ari Berkowitz.
Fruitfly mating and longevity data, from Partridge and Farquhar, via James Hanley and Stanley Shapiro
How to win at roulette!

READINGS FOR BIOLOGY 28 (BIOSTATISTICS), Spring 2000

A. Text (available at the Jeffery Amherst College Store):
Intuitive Biostatistics, by Harvey Motulsky (New York, Oxford University Press, 1995). We will read all but chapters 29, 32, 33, and 34 as part of the course. Chapters 35, 36, 37, and 38 are not mentioned in the course outline, because they cover general subjects such as common statistical errors; we'll refer to those chapters from time to time during the course.

B. Handouts: Many scientific papers and excerpts, to be distributed in class.

Heath, D. An Introduction to Experimental Design and Statistics for Biology, London, UCL Press, 1995. This book contains excellent conceptual explanations of the thinking behind many statistical tests and procedures. Some passages will be provided as class handouts.
Campbell, S. K. Flaws and Fallacies in Statistical Thinking, Prentice-Hall, 1974.
Fisher, R. A. The Design of Experiments. Hafner, 1971.
Gonick, L. and Smith, W. The Cartoon Guide to Statistics. Harper Perennial, 1993
Moses, L. E. Think and Explain with Statistics. Addison Wesley, 1986.
Salsburg, D. S. Understanding Randomness. Lecture Notes in Statistics, Vol. 6 Marcel Dekker, 1983.
Sokal, R. R. and F. J. Rohlf. Introduction to Biostatistics, 2nd 3ed. Freeman, 1987.
Zar, J. Biostatistical Analysis. Prentice Hall, 1984.

My goal is to include evaluations based on situations similar to "real life" use of statistics, i.e. as much emphasis on problem sets and presentations as on exams. Here is the required work of the course:
(1) PROBLEM SETS (6% each, total 36% of the grade): 6 problem sets, involving data presentation, statistical analysis, and experimental design; you will be asked to explain your approach in addition to doing the calculations. Problem sets will be due in class on 6 Thursdays during the semester (Feb. 3 and 17; March 2 and 23; April 6 and 20. For problem sets that you hand in on time (but not those handed in late), you may submit corrections within within 2 weeks for partial credit.
(2) QUIZZES (7% each, total 28% of the grade): There will be 6 in-class quizzes on Fridays (Feb. 4 and 18; March 3 and 24; April 7 and 28). Three of the quizzes will be closed book and three open book. Quizzes will include both conceptual questions and calculations. The best 4 scores will be counted towards your grade, and the lowest two scores will be dropped. Under these circumstances, there will be no makeups for quizzes missed without a letter from the Dean of Students verifying personal illness or family emergency.
(3) PRESENTATION (8% of the grade): Each member of the class either individually or with a partnerwill make a 20 minute presentation to the class, with a few minutes for questions afterwards, on a topic related to scientific method, or cases of interesting use or misuse of statistics. You will provide a 1 - 2 page abstract, and a list of references, for each presentation. Sample topics are listed on an attached pages, but I will be happy to consult with you if you have a suggestion for a different topic.
(4) PAPER (8% of the grade): A 4 - 6 page paper based on your presentation; will be done jointly with your partner if your presentation was a joint presentation.
(5) FINAL EXAM (20% of the grade): A comprehensive, closed book, scheduled final exam
(6) ATTENDANCE AT PRESENTATIONS: attendance is required at the presentation sessions.

Misc

Prerequisites:
There are no college-level prerequisites for this course. Obviously we will constantly use numbers (i.e. calculations) and symbols (i.e. algebra), but the only formal math requirement is at the level of "pre-calculus" courses.

Applying Bio. 28 to your major:
This is a non-laboratory Biology course that counts as a "List A" elective for the Neuroscience major. For Biology majors, Bio. 28 may be used as one of the 5 courses beyond Bio. 18-19 for the major, but it does not fulfill either a lab or an "area"

Topics

Sample Special Project Topics

How dumb can you get!
An article in the December 1999 Journal of Personality and Social Psychology says that dumber people rate themselves as above average, while smarter people say the opposite. Maybe we should look at exactly how the study was done?

The year 2000 census (or should it be a sample?)
The U.S. Constitution says the government has to count the entire population every ten years, and this is one of those years. However, those who know statistics say the count would be more accurate using sampling rather than trying to count everyone - and the errors from trying to count everyone are by no means random.

The Gulf War syndrome
Many veterans of the Gulf War report debilitating illnesses that are unlike any described in medical textbooks. Can statistical methods be used to test what is going on here?

Should women have regular routine mammography screenings for breast cancer?
A recent high-level panel was asked to answer this question, but they couldn't reach a unanimous recommendation. How could something like this not be a good thing?

DNA fingerprinting
How good should a "match" be between DNA at a crime scene and DNA belonging to an accused person in order to convict? Refs.: Risch and Devlin, Science 255: 717-720 (1992); Science 256: 593 (1992); Science 254:1745-1750 (1991)

Smoking and cancer
Is it correlation or causation--or is that even the right question?

Evidence of inheritance of acquired characteristics?
A famous early 20th century case of apparent fraud, complete with sex--in frogs--and the tragic suicide of the scientist involved. Ref: The Case of the Midwife Toad, by Arthur Koestler.

The homeopathic medicine controversy
Played out in the pages of the distinguished journal Nature, a French group (Benveniste et al.) claimed that water has a "memory" of being exposed to certain chemicals that persists even when the chemical had been successively diluted so many times that there should be less than one molecule remaining. This was said to be how some homeopathic remedies work.

The placebo effect
It's one reason why controls are needed in clinical trials. Exactly what is it, and does it have a physiological explanation? Should subjects receive placebos to provide a control in clinical trials when an alternative treatment is available? Science 267: 25-6 (2 Jan. 1995)

Does "oral toleration" using myelin antibodies cure multiple sclerosis?
A recent clinical trial says yes, critics say no.

Facilitated communication in autism
An autistic person who can't speak a word points to letters on a keyboard that spell out meaningful sentences--but only when a "facilitator" gently holds the autistic person's hand. Real communication by the autistic person, or a "clever Hans" effect?

Ape language
Do Sarah the Chimp and Koko the gorilla really use sign language, or are they just going through the motions to get a banana?

Case histories of recent research fraud
Fortunately for students in Bio. 28, but sadly for the integrity of science, there is ample material here for several juicy presentations and reports, from an Office of Research Integrity newsletter from NIH, and from a book by Broad and Wade, Betrayers of the Truth. Also Science 245: 120-122 (1989).

Do chemists die young, statistically?
An article in Am. J. Industrial Med. 23: 615 (May 1993) analyzes the data. Perfect for someone seeking revenge for Chem. 12.

Is schizophrenia a genetic disorder?
Four announcements of finding "linkage" have been made--complete with major press conferences and network TV coverage--in recent years. All four were later retracted, based on new evidence. What's going on? Ref: Alper & Natowicz, Trends in Neurosci. 16 387-9, 1993.

Strategic bombing in World War II
During the London blitz, bombs fell in clusters, sometimes near major military or industrial sites. Statistical analysis was used to analyze whether the Germans actually had the ability to pick out particular targets, or whether the bombing was actually random.

Racism by "statistics"
From cranial capacity to intelligence tests, "science" has sometimes serve to promote belief in racial differences; The Mismeasure of Man, by Steven Jay Gould; The Bell Curve.

The life of "Student"
W. S. Gosset worked for the Guiness brewery. His employers made him publish his pathbreaking statistical work under the pseudonym "Student" because they didn't want the competition to know they were using heavy brainpower to market their product.

The dance of the bees
Does the bees' dance communicate direction and distance to food source, as everyone thought--and introductory Biology and Psychology texts asserted--based on von Frisch's classic experiments? If the answer were "yes", we wouldn't be asking!

Does prayer work?
A recent study apparently shows that coronary care patients for whom someone prayed did better than control patients--and the doctors and nurses taking care of the patients reportedly didn't know who were the "experimentals" receiving prayer and who were the controls, so it was a true "double blind" test.

The frauds of Sir Cyril Burt
Statistical thinking exposed Burt's fake data supposedly showing a high genetic component of IQ differences in "separated identical twins." Ref: Letter in Science 204 242-3.

Genes and IQ
Burt's fraud aside, how good are the concepts and data linking genes and intelligence differences? The Science and Politics of IQ, by Leon Kamin.

ESP
Some demonstrations of alleged psychokinesis, telepathy, and clairvoyance have reported astronomical odds against the outcomes happening by chance. Should we believe?

Animal "super-senses"
Perhaps more plausible than ESP are reports of animals' ability to detect things humans can't, e.g. that an earthquake is about to happen, or that their epileptic owner is about to have a seizure. Still, there is no explanation of how this could be possible. (Springfield Union News, 17 Aug. 1995).

Psychology of probability
People tend to make systematically skewed judgements of probability; there is a whole subfield of Psychology to investigate this bias. Judgment Under Uncertainty, by Kahneman, Slovic, and Tversky.

Feminist science
There are many potential topics here, some more relevant to Bio. 28 than others. The misuse of statistics to justify sexism, and the claim that women scientists approach science differently, would be good subjects for this course.

A statistical theory of some kinds of cancer
A "two-hit" model fits very well the data on incidence of retinoblastoma in one eye vs. both eyes. Ref: Cevence and White, Sci. Am. March 1995, p. 73.

Attacks on misuse (and perhaps use) of statistics
.. and not because they're boring, either! Refs: Tainted Truth, reviewed in NYTimes 31 July 1994; a Sci. Am article on "The Anti-Scientific Attitude".

Can there be a statistical measure of the value of a human life?
A smoke detector reduces the probablity of dying in a fire by 1 chance in 10,000. People who will buy one for \$20 but not for \$50 seem to be placing a value on their own life of between \$200,000 (\$20 x 10,000) and \$500,000. Government regulations prohibiting using asbestos in auto parts cost consumers extra money and save only a tiny number of lives; the extra cost per life saved is \$1.2 billion. Is the regulation against asbestos justified? Is this kind of calculation rational in the first place? Refs: NYTimes 29 Jan. 1995; Science 248: 559-564, 1990.

Why are spouses good organ donors?
Kidney transplants between spouses (who of course are genetically unrelated) are almost as successful as transplants between identical twins (NEJournal of Medicine, 8 August 1995). Why so?

When randomly chosen people are polled, on the average they rate their own risks of various hazards (flu, cancer, poisoning, etc.) as below average. (Perhaps it's the opposite of the "Lake Wobegone Effect", where according to Garrison Keillor all the children of the town are above average.) Can anything be said in favor of this optimistic bias? Ref: N. Weinstein, Science 246: 1232-3, 1989.

Human behavior and relative risk
People who fear airplane flight regularly put their life in jeopardy driving, which statistically is much more dangerous than flying. Others who protest nuclear power plants put themselves at much higher risk of death from heart disease by eating fatty foods and not exercising. Are people just irrational, or what?
(NY Times, 6/15/97)

Vaccination risks
When do the risks of vaccination exceed the risks from the disease that is prevented by vaccination? Should an individual give the same answer to this question as a public health official would? (Nature 318: 323-9, 1985; Lancet 335: 641-5, 1990).

Are some infinite, non-repeating decimal numbers more random than others?
By one newly developed test for randomness, pi was more random than e, which in turn was more random than the square root of 2. (Science 276:532, 4/25/97)

The draft lottery wasn't random
As Amherst Prof. Norton Starr has written, the 1970 draft lottery, on the basis of which many young American men went to war in Vietnam and some died, was supposed to choose draftees at random based on birthdates. However, it didn't turn out that way. (http://www2.ncsu.edu/ncsu/pams/info/jse/v5n2)

Benefits of sex
"Study of men finds regular sex may reduce risk of death by half." Whether this excuse is needed or not, is it true? (Springfield Union News, 12/23/97.)

Violence and the weather
"Study proves it: Heat makes people violent." Can it be so? Springfield Union-News, 11/27/97)

Magnetic fields and cancer
For several decades, there have been reports that people exposed to low-frequency electromagnetic fields have increased risk of some kinds of cancer. What's the evidence? (Science 277: 29-30, 1997)

Non-random "randomized" trials
Evaluation of new drugs and therapies in medicine depends on randomized clinical trials. Recently it has come to light that some supposedly randomized controlled trials weren't random at all. ( JAMA 274: 1456-1458, 1995.)