DATA REDUCTION WITH GROUPING AND WEIBULL MODELS.
Interim scientific rept.,
MICHIGAN STATE UNIV EAST LANSING DIV OF ENGINEERING RESEARCH
Pagination or Media Count:
The computational details are given for a two-step procedure for condensing size-graded data, especially bio-medical data, into a parametric statistical model. The first step establishes a histogram of size density while the second fits a Weibull model to the histogram. Three types of histograms are defined and compared on the basis of computational efficiency. An algorithm is proposed for simultaneously estimating the scale, shape, and location parameters of a Weibull distribution from a histogram, and figures of merit for the Weibull fit are defined. The Weibull model is shown to be more flexible than an Edgeworth model. Examples are provided based on computer-simulated data and electroencephalographic EEG data. One example classifies EEG sleep data according to sleep stage with the shape parameter of a Weibull fit to a frequency histogram. Author
- Medicine and Medical Research
- Statistics and Probability
- Computer Programming and Software