|
Activity Number:
|
55
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Sunday, July 29, 2007 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
IMS
|
| Abstract - #309397 |
|
Title:
|
Spatial Statistical Analysis of Doctors' Prescription Amounts by Region
|
|
Author(s):
|
Lei Kang*+ and Noel Cressie and Desheng Liu
|
|
Companies:
|
The Ohio State University and The Ohio State University and The Ohio State University
|
|
Address:
|
Department of Statistics, Columbus, OH, 43210-1247,
|
|
Keywords:
|
EDA ; ESDA ; CAR model ; Maximum likelihood estimation ; Spatial dependence
|
|
Abstract:
|
In this talk, we analyze doctors' prescribing patterns based on regional aggregation of individual prescription data. Through a series of exploratory-data-analysis (EDA) techniques, we demonstrate the importance of transforming the data and of doing a weighted regression analysis. Both a non-spatial model and a conditional autoregressive (CAR) model are fitted. Our results show that CAR models have smaller sum of squared prediction errors compared to non-spatial models. Different decompositions of large- and small-scale variation are considered; we see that when a lot of spatial variation is attributed to the large-scale, deterministic component, spatial-dependence parameters describing the small-scale, stochastic component decrease in value with larger standard errors. This work motivates the use of a hierarchical spatial statistical model to describe doctors' prescribing patterns.
|
