SOME TEACHING DATA SETS

VA hospital quality monitor
The outcomes in the monitor areas represent annual numbers of individuals who failed to return for an outpatient visit within 30 days of discharge, out of the total number of annual discharges. Monitor 21 here is in the Substance Abuse Psychiatric care. Low return rates are indicative of low ``quality'' in these specific care areas. This subsample of a much larger data sets provides information on years 1992 and 1993 for the 159 hospitals in the VA system.
Here are the same two years of outcome data, but now each hospital/year combination has an additional number associated with it: the VA's case-mix adjustment factor that, for each hospital in each year, is supposed to be a predictor of the fraction of returns. This plays the role of a covariate in models for the outcomes. The two years are given here separately, 1992 and 1993
Pollutant levels
Recordings of levels of a photochemical pollutant (some kind of oxidant) made hourly at several recording stations in LA (by the LA Polution Control District). Interest lies in the relationship of pollutant levels to one or more meteorological variables.
Used car prices
Some data I once collected on asking prices of used Mercedes, classified by several factors.
Will you take mercury with your fish?
Data from a recent study conducted in the Wacamaw and Lumber Rivers (NC) to investigate mercury levels in tissues of large mouth bass, of obvious interest in connection with environmental health as such fish are a regional delicacy (a concentration over 1 part per million).
Old Faithful geyser eruptions.
A classic data set giving recorded times between consecutive eruptions of the famous Old Faithful hot water geyser in Yellowstone National park. An interesting problem: predict the time of the next eruption. This data is 20 or so years old, but Old Faithful's habits don't change much. Here's some more uptodate info
Neural function
Data measuring the induced levels of electrical potential in a mammalian CNS nerve cell following stimulus to "excite" neural transmission at a synapse. This data arises in investigations of the fundamental stochastic nature of electro-physiological communication in nervous systems, at the subcellular level. These data come from one of many similar experiments in the Duke Medical Center and Durham VA Neurosurgery departments.
EEG (electroencephalogram) recordings.
Here is a long series of 3600 EEG recordings from a long EEG trace recorded in the ECT Lab at Duke, on a patient undergoing ECT therapy for clinical depression. Here is a short segment of only 400 observations from the central section of the EEG series. The data are fluctuations in electrical potential at a point on the patient's scalp during seizure, one of several "channels" of recordings. They are measured in microvolts and represent measurments taken at time intervals of roughly one fortieth of a second.
Abalone shell fish measurements
A data set from the Marine Resources Division at the Department of Primary Industry and Fisheries, Tasmania. Recorded measurements on a few thousand Abalone, of interest in determining relationships useful to predicting the age of future abalone from easily made physical measurements. The age of abalone is determined by cutting the shell through the cone, staining it, and counting the number of rings through a microscope -- a boring, time-consuming and expensive task. Other measurements, which are easier to obtain, are therefore used to predict the age. (From the original data examples with missing values were removed and the ranges of the continuous values have been scaled by dividing by 200). Why is it interesting to predict age? For ecological and commerical fish farming purposes, the age composition of abalone populations are relevant. Here the number of rings is proxy for age.
HIV babies 1999 update
A February 2, 1999 New York Times report on a recent follow-up study of HIV positive mothers and newborns. This discusses results that extend the conclusions we make from the older (Feb 1998) data -- referred to here as the "Thailand study" -- and discusses issues of ethics in clinical trials (as discussed in the Problems of Proportions handout)
Yen/US exchange rates and interest rates
Monthly data on: Japanese Yen short interest rates (JPYSI), US interest rates (USDSI), and Yen/USD exchange rates (JPYUSD) from July 1978 to December 1998
Exchange rates
Changes in the daily spot exchange rate of the US dollar to the British pound over a period of 1000 days ending on 8/9/96.
Here is a much larger exchange rate data set. Daily spot prices and corresponding returns for several years.
Tamoxifen: Breast Cancer Drug Trials
Data from 14 clinical trials that studied the effects of the cancer drug Tamoxifen on women with early breast cancer. Each study delivers a 2x2 table of outcomes comparing two groups: treatment and control.
Sex bias in salaries?
A subsample of a large data set (from the early 1980s) from a study investigating potential sex effects (biases?) in determination of professional salary differentials. The individuals come from several large corporations.
Southern oscillation index
Monthly values of the The Southern Oscillation Index (SOI) of interest in connection with warming trends and "El Nino" cyles in cimate.
Oxygen isotope levels
Measures of relative abundance of Oxygen-18 to Oxygen-16 over a period of about 1.5million years, of interest in studies of climate patterns and climatic change, and the effects of the orbital dynamics of the Earth
US college tuition levels
A large data set recording annual tuition levels of over 1000 US colleges, together with various other variables that may be of use in describing variation in tuition levels.
GDP time series
Annual per capita GDP time series for several countries. One interest lies in studying the "periodic" behaviour of such series in connection with understanding business cycles. Another lies in forecasting turning points. Another lies in comparisons of characteristics of the series across national economies.
Nasopharynx cancer
Survival times of patients with identified nasopharynx cancer tumours, with information about age at diagnosis, treatment type, sex and other variables. Questions of interest relate to the form of survival distributions and how treatment, age, extent of tumour growth and other factors impact survival times.
More breast cancer
Survival times of breast cancer patients under two forms of treatment. Questions of interest relate to the form of survival distributions and comparison between two treatments (one standard, one experimental). Here is some S-Plus code to explore this data using two-sample exponential models.