CJPIV/SR2022Malnutrition

Colgate University, Ay Lab - ML Analysis of undernutrition in Ethiopian schoolchildren # SR2022Malnutrition Ay Lab, Colgate University\ Summer 2022\ Jim Perry, Will Russel This is an open-source data science script developed by research students at Colgate University, designed to employ various machine learning and statistical methods to analyze a 2013 dataset on the prevalence of three types of malnutrition (stunting, thinness, underweight) in Ethiopian schoolchildren. These methods include imputation of missing/unanswered survey data through the kNN algorithm (and cross-tested with RandomForest imputation), one hot encoding of categorical variables, various feature selection methods (Multivariate Logistic Regression, Chi-Square test, mRMR, JMI, Monte-Carlo), further linear and logistic regressions for odds ratio calculations, and finding rules in data through both the Classification and Regression Trees (CART) and the Apriori Assocation Rule Learning algorithms. These analyses are carried out on both general data and selected subsets of interest: these include Male/Female, Urban/Rural, and Youth/Adolescent (respectively >10 and >=10 years old). This subset analysis was performed on groups of positive samples for the three types of malnutrition tested. Original survey data is available upon reasonable request. *INSTRUCTIONS FOR RUNNING:* 1. Install RStudio and/or make sure it's up-to-date. 2. Download and unzip this repository. 3. Open "Malnutrition Masterscript.Rmd" in RStudio. 4. Run "Malnutrition Masterscript". You can press ctrl+shift+R to run everything at once. Alternatively, you can click on specific chunks and use ctrl + shift + enter to run the entire block. Necessary packages are automatically installed at this step, as they're the first segment of code run. 5. (OPTIONAL) Adjust number of features selected on line 931. Default is 10. 6. Run "Malnutrition Masterscript". You can press ctrl+shift+R to run everything at once. Alternatively, you can click on specific chunks and use ctrl + shift + enter to run the entire block. *ACCESSING RE …