Event Title

Forecasting for the All Women Count! Program

Location

University Student Union: Volstorff A

Start Date

12-2-2018 12:00 PM

Description

All Women Count! program is a no-cost breast and cervical cancer screening program for qualifying women. We are interested in estimating the number of women who will use the program for the next 5 years. Forecasting was done using several commonly used models for each county. In addition, a Gaussian mixture of regression time series model is used to perform clustering and forecasting. Four models were tested and the model with the lowest test root mean square error was chosen to carry out the forecasting by county. The model chosen most often was the ordinary least squares regression closely followed by ridge regression and linear regression with autoregressive integrated moving average errors. Model selection for the mixture model was done using the Bayesian information criterion and found 5 clusters were optimal. The five clusters identified the counties with increasing and decreasing participation. The results will help the South Dakota Department of Health with future planning and implementation of the program.

This document is currently not available here.

Share

COinS
 
Feb 12th, 12:00 PM

Forecasting for the All Women Count! Program

University Student Union: Volstorff A

All Women Count! program is a no-cost breast and cervical cancer screening program for qualifying women. We are interested in estimating the number of women who will use the program for the next 5 years. Forecasting was done using several commonly used models for each county. In addition, a Gaussian mixture of regression time series model is used to perform clustering and forecasting. Four models were tested and the model with the lowest test root mean square error was chosen to carry out the forecasting by county. The model chosen most often was the ordinary least squares regression closely followed by ridge regression and linear regression with autoregressive integrated moving average errors. Model selection for the mixture model was done using the Bayesian information criterion and found 5 clusters were optimal. The five clusters identified the counties with increasing and decreasing participation. The results will help the South Dakota Department of Health with future planning and implementation of the program.