MISY262 Lecture Notes - Lecture 25: Data Model

• School: University of Delaware
• Department: Management Information Systems
• Course: MISY262
• Professor: Kathryn Berkow

Document Summary

Data model: fit, fit, fit training data error distribution. Observations occur over time: uses latest data as test set, use: train = 1:(. 7*n) the first 70 numbers. Observations occur over people? items: randomly select test set, use: train = sample(1:n,. 7*n,replace=false) Use prediction to assess logistic regressions fit statistics do exist. How should we divide the data? create model: sleep ~ grade + hours log(p/1-p) = -7. 10 + . 10*grade p = probability student got sleep on the night before the exam. Predict: did student get sleep if grade = 85? log(p/1-p) = -7. 10+. 10*grade: calculate k: k = -7. 10 + . 10 * 85 = 1. 4, calculate p: p = e1. 4/1 + e1. 4 = . 802.