lmaesampson/CNSCat

Determining the cause of central nervous system (CNS) infections in kids in Uganda using clinical data at intake. # Project Overview: Given clinical data from approximately 400 children in a Ugandan hospital, we explore if a trained classifier can improve diagnostic accuracy above that of trained experts. We look at decision trees and Bayesian logistic regression as two methods to compare. We also explore this problem as one of assessing different operational strategies. Is it 'best' to follow doctor's best guesses, the output of our classifier, or simply to treat all of the children with all available medications at intake? **Contributors** Laura Sampson, post-doc at CIDD at Penn State University Michael Williams, graduate student at Vanderbilt University # Folder Contents: **Project Def**: Statement of problem, raw data, definition of variables in raw data, etc. **Michael**: All of Michael's analysis. This include *Logistic Regression*: Logistic regression as implemented by Michael *Exploratory Analysis*: Michael's initial passes at working with the data and implementing regression **Laura** *Code*: Laura's various codes - decision tree, data cleaning, logistic regression, assessment of operational strategies. *Data*: Data files with some initial cleaning completed.