Appendix 1¾Syllabus ©2003, D.F. Parkhurst
“Most real-life statistical problems have one or more nonstandard features. There are no routine statistical questions; only questionable statistical routines.” – D.R. Cox
"…
much of what statistics teaches seems surprising, counterintuitive, and even
unscientific."
– R. Royall
E538¾Statistics for Environmental Science¾Spring 2001
School of Public and Environmental Affairs¾Indiana University
Professor: David Parkhurst Section 8442
Office: SPEA 355 9:30–10:45
a.m. MW, SPEA 273
Phone: 855-4556 1–3
p.m. F, SPEA 277
E-mail: parkhurs@indiana.edu
Office Hours: Monday 11:00-12:15a Section 8444
Tuesday 1:30-2:30p 2:30–3:45 p.m. MW, SPEA 277
and
by appointment other times 9–11
a.m. F, SPEA 277
Associate
instructor: To be announced
Purpose of
statistics
The
primary purpose of statistics is to allow us to use data to improve our knowledge,
and, in applied situations, to use that knowledge effectively in making
decisions about how to act. As
illustrated here (and as will be described further in class), that process
requires knowledge both of the subject being studied, and of statistical and
data-analysis methods. The latter are
the focus of this course.
Course
description:
The popular view of “statistics” seems to correspond
with the title of the infamous book, How
to Lie with Statistics (D. Huff. 1954. Penguin.), but a worthy alternate goal for environmental scientists is to
use statistical methods to help them learn what is true, and to act
appropriately as a result. We need
tools for analyzing data, with their associated variability and uncertainty, to
improve our understanding (basic science) and to guide our management decisions
and actions (applied science).
To analyze data effectively, environmental
scientists need to know some basic statistical “recipes,” but also need to know
when those recipes should or should not be used, and how to interpret the
results they produce. Thus the concepts
that underlie the recipes are important as well. Beyond this, statistics is currently undergoing several major
revolutions that demonstrate both progress in the field and real problems with
the conventional statistical methods that most scientists now use. In addition to conventional statistics,
therefore, we will consider three of these revolutions at least briefly, i.e.:
· Bayesian methods, which are
based on so-called subjective or personal probabilities, and which often give
answers much closer to what scientists really want to know, compared with most
answers available from “standard statistics,”
· Elimination of, or greatly
reduced emphasis on, null hypothesis significance testing, which has many
logical problems and tends to hide adverse environmental effects, and
· Resampling methods, which
reduce the need to make sometimes dubious assumptions about data.
We will also consider what these alternative views of data analysis teach us about the shortcomings of conventional methods.
Our overall approach should help you to be effective
environmental scientists and managers, not only in the short run but also as
your career develops in the longer term.
Course
objectives:
The three major goals of this course are that you
should:
1.
Learn
how to separate the "signal" (or information, or pattern) from the
"noise" (or random variation) in data, and then to use the
information obtained to increase your knowledge and when appropriate, to help
your decision making.
2.
Learn
how to obtain data in efficient ways, to provide maximum information for the
time and money spent.
3.
Learn
how to recognize illogical uses and interpretations of statistical methods by
others, to aid in advancing scientific knowledge and for effective environmental
management.
I will emphasize conceptual understanding with the
hopes that in the future:
· You will know when standard
statistical methods are being properly applied and when they are not.
· You can modify and extend
standard (and non-standard) methods to fit new situations.
For each major topic, I will try to follow a common
pattern by discussing:
· What the concept is, and why
it is useful.
· How to do any necessary
calculations (sometimes by hand and sometimes with software).
· How and why the concept works
(simple theory).
· Examples from environmental
science, from the viewpoint of both basic and applied science.
Course format and resources
We deal with complex but important ideas in E538, and many students will have to invest considerable effort in the course. We will help you as much as we can through:
· Lectures. Most days, about
an hour of the class will be in lecture mode. You should feel free to interrupt
and ask questions if you “lose the thread.”
· In-class problems. I will
frequently pose problems for you to work, preferably in groups of two or three.
We’ll discuss these after you’ve had some time to attack them.
· Discussion. Occasionally,
lecture time may be replaced with small group discussion, to help you grapple
with important philosophical issues.
· Textbooks and notes. The Moore and McCabe book is an introductory
but substantial text that is appropriate for a strong undergraduate or
introductory graduate course in statistics.
I have chosen it because it gives a solid basis for the more advanced
consideration we give to certain topics that practicing environmental
scientists ought to know. (Parts of
this book will be review, but some of the material will be new for most
students. In either case, you can
expect to see at least a few exam problems based on material from the book that
is not repeated in lecture.) Advanced
topics are covered in lectures and in the supplementary lecture notes that you
should pick up at a local copy shop¾listen for an announcement
in lecture. One such topic is Bayesian statistical
methods; the Iverson book provides a
readable entry into that subject.
· Labs. These Friday sessions
allow time, among other things, for sampling exercises in which each student
helps generate part of a large dataset to illustrate some critical point. In
addition, the lab periods encourage you to work together on problems. The teaching assistant will be there to
provide examples and to help with the assigned problems.
· Exercises. You will be
provided with sets of exercises each week to show how the principles under
study apply to environmental science and management. Many of these problems are
taken from past exams, so they give you practice solving the kinds of problems
you can expect to find on the exams.
Study hint: We will provide detailed answers for some problems, but we
strongly recommend that you not look at the details until you have worked the
problem yourself. You will learn a thousand times more (that’s a rough
estimate) by working a problem through yourself than by looking at a prepared
solution and convincing yourself that you understand it. Remember that in your future work you will
have to solve problems yourself, not just understand someone else’s solutions,
and the same will be true in exams.
If you
help students with a problem, you can do them the most good not by working it
for them, but rather by either (a) working some similar problem for them, or
(b) finding out where they are “stuck,” and asking leading questions that will
get them on the right track. If you go
to the effort to help others in these ways, you will likely learn a lot
yourself in the process.
· Exams. As a student, I learned a great deal from
the process of wrestling with ideas while taking exams, and I try to write
questions that will give you that same benefit. This means that most questions will test your knowledge of
underlying concepts, and not just your ability to “spit back” standard
recipes. Questions will be similar in
general form and difficulty to the assigned exercises and to those on past
exams, but will not be identical to them with just the numbers changed.
· E-mail. Each day at the end of class, I will ask
part or all of the class to write out a brief note about (a) what you found
most interesting that day, (b) what you understood least that day, and (c) any
specific questions you might have.. I
try within a day or two to provide answers to questions, via e-mail that I send
to everyone in the class. PLEASE READ
THESE MESSAGES—I view the information they provide as supplements to lecture,
and will assume you have that information when I write exam questions.
· Office hours: Both Professor Parkhurst and the associate
instructor hold regular office hours, and will be happy to help you when you
can’t work a problem or don’t understand some concept. See Dr. Parkhurst after class or send e-mail
to make appointments with him for other times.
· Homework: Seven or eight homework assignments spread
through the semester will give you practice in analyzing data and interpreting
statistical results.
Grading:
Many exam and homework problems involve sequences of calculations and logical steps. When we find an error in your work, we try to count off once for that error, and then to grade the rest of the problem starting from that point. That is, we try not to let a small error at the start of the problem cause you to lose too many points overall.
After each exam, I will give you the conversion from
scores to grades, so you know right away the contribution of that exam to your
course grade. I decide on the
conversions by the following sequence of reasoning. First, I treat each score as a percentage of the points
available, and assign grades on a scale where 90% = lowest A-, 80% = lowest B-,
70% = lowest C-, etc. My exams can be
challenging, and if that process leads to too many low grades, I will boost the
distribution, usually with a conversion of the form 
where a and b are appropriate constants.
Texts:
· D.S. Moore & G.P.
McCabe. 1998. Introduction to the Practice
of Statistics (3rd ed.). W.H. Freeman & Co.
· Class notes. To be purchased at a local copy shop.
· G.R. Iverson. 1984. Bayesian
Statistical Inference. Sage Publications. Newbury Park, CA.
· You will need a pocket
calculator with log x, ex, å, and regression
functions. Calculators with several
internal memories help you avoid writing down intermediate results. A single storage register is acceptable,
however.
· This may be useful for
some: Gonick, L. and Smith, W. 1993. The
Cartoon Guide to Statistics. New York, Harper Perennial. This is a well written book that might be
helpful to you. You can probably find it in local bookstores¾I have
not ordered copies for the course, however.
· Please bring graph paper to
every lab. For consistency, I recommend
paper with one-cm squares, and lighter one-mm squares.
Prerequisites:
This course has a prerequisite of an introductory
course in statistics, and that background will
be assumed. It is possible you may
have learned the necessary concepts in some science class, but if you haven’t
specifically had a stats course, please see me right away.You should also have
a firm grasp of freshman level algebra and of introductory calculus, including
both derivatives and integrals.
We will encourage you to perform (or sometimes to
check) your homework calculations using statistical software like SPSS (or
perhaps other packages).
Grading:
The two mid-term exams will each contribute 28% to
your course grade, and the final exam, which is comprehensive, 34%. The homework will contribute the remaining
10%.
Course ethics:
Many of my colleagues share with me the frustration
that a few students seem increasingly not to know the difference between
intellectual honesty and dishonesty.
Cheating on tests is obvious enough, but issues involving out-of-class
assignments (like the homework for this class) seem less clear. These difficulties are increased by my hope
that you will work together¾to a point¾but that you will turn in
work that is your own. Computers add
another source of ambiguity. The
following discussion is intended to help clarify my expectations about the
homework. Please ask if you want
further guidance.
Here is a statement (modified from one provided by Prof. D. Willard) that I use in my undergraduate classes:
ACADEMIC MISCONDUCT. Academic dishonesty is not common, but is serious and intolerable.
Please familiarize yourself with the Student Handbook guidelines. I assume that you all know and understand what plagiarism and cheating
are; if you don't, find out. The rules are simple. Do your own work. Don’t copy or even seem to copy from others. Allowing someone else to use your work as if
it were their own is as serious as using someone else's work without full
written acknowledgment in whatever you turn in. If you make legitimate use of work done by others, always
document your sources.
For E538 homework, I hope that you will work together in the following
acceptable ways:
· Verbally discuss the
purposes of an assignment. What is the point
of an analysis like this? Why would we
want to know the results being asked for?
What assumptions are required for a particular kind of analysis? Why are these methods useful? In what ways is the exercise like something
you have done in the past, or might have to do in the future? What would the results mean if they came out
in various ways?
· You may also discuss the
general ways of dealing with the types of calculation required for the
exercise. If you know how to do the
calculation, and want to help a classmate who hasn't figured it out yet, I
would be pleased if you did that.
Indeed, probably the best way to learn well how to do something is to
show someone else how to do it¾ such cooperation should
help both you and the person you are helping.
However, do this by making up a similar set of data and showing your
classmate how to analyze the made-up data.
Then, leave them to analyze the
real homework exercise themselves.
· Similarly, if you are
helping someone learn to perform statistical analyses (for example, in SPSS),
show them the general ideas, but do not leave a computer file in a state such
that they can enter the homework data into a ready-made template that you have
prepared.
Here are some examples of what is NOT permissible:
· Copying from another
student's written answers, even if you make substantial changes to the
wording. Also, as noted above, it is
just as unacceptable to provide your answers to another student as it is for
them to use your answers.
· Analyzing the data in a
computer file, then allowing someone else to use or modify that file (or a copy
of it) in any way.
· Giving a copy of any computer file related to the
homework assignment to any other student in the class. The one exception to
this is that if two or more students are to analyze the same data, one person
may type in the data and pass a file containing
the raw data only to others.
If you are uncertain about what is fair and what is
not, please err on the careful side, then ask for clarification later.
Policy on
“Incompletes”:
The University policy
on grades of "incomplete" includes the following statement:
CIRCUMSTANCES PERMITTING INCOMPLETES
The grade of Incomplete used on the final grade
reports indicates that the work is satisfactory as of the end of the semester
but has not been completed. The grade
of Incomplete may be given ONLY when the completed portion of a student's work
in the course is of passing quality.
Instructors may award the grade of Incomplete upon a showing of such
hardship to a student as would render it unjust to hold the student to the time
limits previously fixed for the completion of his or her work.
After discussing this
statement with my colleagues, I believe that the "hardship" referred
to does not include poor preparation or planning, an overloaded schedule, or
similar factors. Rather, it refers to
substantial illness, family emergencies, and the like. Any incompletes granted in E538 will be
based on this University policy.
Policy on exams:
You must take mid-term
exams at the times they are scheduled for the lab in which you are
enrolled. (I may, however, attempt to
arrange for the exams for both sections to occur at a common time.) If that creates a conflict with some other
course, you may ask to take the exam for the other lab section and I'll most
likely give you permission to do so, depending on availability of space. Such requests will be considered only if
made at least two weeks prior to the exam time.
You must take the
final exam at the University's published time (extended slightly as listed just
below) for the section in which you are officially enrolled. It is your responsibility to arrange
employment requirements, job interviews, airplane flights, and the like so they
do not conflict with any of the scheduled exams. The times, as extended a bit from page 28 of the semester's
schedule of courses, are:
Section 8442 (9:30
a.m. lecture)¾Final
exam from 9 a.m.– noon on Wednesday, May 2.
Section 8444 (2:30
p.m. lecture)¾Final
exam from 10:15–1:15 p.m. on Monday, April 30.
E538—Spring 2001—Tentative Schedule:
Week of
|
Topic |
Reading[1] |
|
Jan
8 |
Introduction,
Distributions, display, and description of data; Pre-test (optional) |
Preface,
Introduction 1 |
|
Jan
15 |
M.L.
King Day (No class Jan 15) Relationships
among two or more variables |
2
(omit 2.6) |
|
Jan
22 |
Experimental
Design |
3 |
|
Jan
29 |
Probability |
4; Iv7–9 |
|
Feb
5 |
Probability,
conditional probability |
4.4; Iv12–17 |
|
Feb 12 |
Binomial and Poisson distributionsProperties
of sample means Introduction
to estimation |
5.1,
Notes 5.2 6.1 |
|
Feb 16 |
First mid-term exam, in
lab |
|
|
Feb
19 |
Decision
making under uncertainty---I Decision
trees, probabilities, and values |
Notes |
|
Feb
26 |
Decision
making under uncertainty---II Randomization
hypothesis tests Types
I and II error, |
Notes 6.2–6.4 |
|
Mar
5 |
Decision
making under uncertainty---III t
tests |
Notes 7.1,
7.2 |
|
Mar
12 |
Spring
break! |
|
|
Mar
19 |
t
tests and power |
|
|
Mar 26 |
Inference
about variability Inference
about proportions |
7.3 8.1 |
March 30 |
Second mid-term exam, in
lab |
|
|
Apr
2 |
Bayesian inference for proportions and for means |
Iv9–11,
Iv18–39, Iv70–77 (Review
earlier Iverson) |
|
Apr
9 |
Transformations |
Notes |
|
Apr
16 |
One-way
anova Two-way
anova |
12,13 Notes |
|
Apr
23 |
Regression
statistics, Logistic regression; |
10,
15 |
|
|
Final exams—time and place as above |
|
