Reorganization

Author: LProcopi15

Week 4 - Case Study/Week4_mortality.R

@@ -2,7 +2,8 @@
 #############Medical decision making: What contribute(s) to mortality?############
 #############Step 1: Understand the dataset###########
 #1.1 Load the data
-
+MD_medical <- read.csv("C:/Users/Student/Documents/UVA 2016-2017/RWorkshop/Week 4 - Case Study/MortData_medical.csv")
+MD_personal <- read.csv("C:/Users/Student/Documents/UVA 2016-2017/RWorkshop/Week 4 - Case Study/MortData_personal.csv")
 
 #Merge two datasets
 colnames(MD_personal)[1] <- "Patient_id"
@@ -13,6 +14,7 @@ View(MD)
 
 #1.2 Read the description files and understand what is each attributes.
 #Rename attributes if you find it helpful
+colnames(MD) <- c('Length.Of.Stay','Disease.Class','Comorbidities','Coma.Score','Care.Intensity','Mean.BP','WBC','Heart.Rate','Temperature','Blood.Gases','Albumin','Bilirubin','Creatinine','Sodium','Adjusted.Shock.Index','Age','Sex','Race','Death')
 
 
 #1.3 Check the class of each variables. 
@@ -34,8 +36,7 @@ View(MD)
 
 
 #2.2.1 Temperature: mean imputation
-
-
+mean(MD$Temperature, na.rm = TRUE)
 
 
 #Additional question: why can we use mean imputation for temperature? Consider the disbribution of temp attribute.
@@ -49,7 +50,10 @@ View(MD)
 #2.2.3 pafi: Blood Gase
 #Consider the real meaning of the attributes. Let's convert the attribute into a factor with two levels, ventilator patients and non-ventilator patients
 #Hint: you can use cut function. Type ?Cut to see the usage of the function
-
+# ventilator patients non-ventilator patients
+MD$Blood.Gases[which(is.na(MD$Blood.Gases))] <- 0
+max(MD$Blood.Gases)
+MD$Blood.Gases <- cut(MD$Blood.Gases, c(0, 32, 870))
 
 
 #In medical world, bili and albi are measured in the same test. 

FireInfo …ek 5 - Introduction to Modeling/FireInfoFireInfo renamed to Week 5 - Introduction to Modeling/FireInfo FireInfo renamed to Week 5 - Introduction to Modeling/FireInfo

@@ -1,74 +1,74 @@
-fire <- read.csv("C:/Users/Student/Documents/forestfires.csv")
-View(fire)
-
-#####################
-#
-# Explore Data
-#
-####################
-
-# Get summary information
-summary(fire)
-# No NA's, all numeric data is normalized
-
-# Check class of each attribute
-for (i in 1:ncol(fire)){
-  print(paste(colnames(fire[i]), ": ", class(fire[,i]),sep = ""))
-}
- 
-# Getting a subset of the data
-summary(fire$area)
-areabox <- boxplot(fire$area)
-upperwhisk <- areabox$stats[5,]
-
-xtm_fire <- subset(fire, area >= upperwhisk, select = c('FFMC', 'DMC', 'DC', 'ISI', 'temp', 'RH', 'wind', 'rain', 'area'))
-
-####################
-#
-# Basic modeling - Linear Regression
-#
-####################
-
-# Plot two variables
-plot(xtm_fire$temp, xtm_fire$area) 
-# Create linear regression model between the two
-lm1 <- lm(area~temp, data = xtm_fire)
-# Add regression line to plot
-abline(lm1, col = "orange")
-# Statistical information about this lm
-summary(lm1)
-
-# Can add many more factors to this lm
-lm2 <- lm(area~temp+FFMC+wind, data = xtm_fire)
-summary(lm2)
-
-# Could interactions between variables help us?
-pairs(xtm_fire[c('FFMC', 'DMC', 'DC', 'ISI', 'temp', 'RH', 'wind', 'area')])
-symnum(cor(xtm_fire[c('FFMC', 'DMC', 'DC', 'ISI', 'temp', 'RH', 'wind', 'area')]))
-# Highly correlated: DC and DMC, ISI and FFMC, temp and FFMC
-# Add these correlated attributes to a model
-lm3 <- lm(area~temp+FFMC+wind+(DC+DMC)^2+(ISI+FFMC)^2+(temp+FFMC)^2, data = xtm_fire)
-summary(lm3)
-
-# compare models
-anova(lm1, lm2)
-anova(lm1, lm3)
-anova(lm2, lm3)
-
-#####################
-#
-# Practice Problems
-#
-####################
-
-# 1. Create a new subset that includes only with an ISI (inital spread index) greater than the median
-
-# 2. Create a linear model with area and wind as your predictors, and area as your response
-# Call this model lm1_ISI
-
-# 3. Determine if there are any correlated attributes
-
-# 4. If there are any correlated attributes add the interaction between them to a new model 
-# Call this model lm2_ISI
-
-# 5. Compare these two models and determine which model is better at predicting the size of the area burned
+fire <- read.csv("C:/Users/Student/Documents/forestfires.csv")
+View(fire)
+
+#####################
+#
+# Explore Data
+#
+####################
+
+# Get summary information
+summary(fire)
+# No NA's, all numeric data is normalized
+
+# Check class of each attribute
+for (i in 1:ncol(fire)){
+  print(paste(colnames(fire[i]), ": ", class(fire[,i]),sep = ""))
+}
+ 
+# Getting a subset of the data
+summary(fire$area)
+areabox <- boxplot(fire$area)
+upperwhisk <- areabox$stats[5,]
+
+xtm_fire <- subset(fire, area >= upperwhisk, select = c('FFMC', 'DMC', 'DC', 'ISI', 'temp', 'RH', 'wind', 'rain', 'area'))
+
+####################
+#
+# Basic modeling - Linear Regression
+#
+####################
+
+# Plot two variables
+plot(xtm_fire$temp, xtm_fire$area) 
+# Create linear regression model between the two
+lm1 <- lm(area~temp, data = xtm_fire)
+# Add regression line to plot
+abline(lm1, col = "orange")
+# Statistical information about this lm
+summary(lm1)
+
+# Can add many more factors to this lm
+lm2 <- lm(area~temp+FFMC+wind, data = xtm_fire)
+summary(lm2)
+
+# Could interactions between variables help us?
+pairs(xtm_fire[c('FFMC', 'DMC', 'DC', 'ISI', 'temp', 'RH', 'wind', 'area')])
+symnum(cor(xtm_fire[c('FFMC', 'DMC', 'DC', 'ISI', 'temp', 'RH', 'wind', 'area')]))
+# Highly correlated: DC and DMC, ISI and FFMC, temp and FFMC
+# Add these correlated attributes to a model
+lm3 <- lm(area~temp+FFMC+wind+(DC+DMC)^2+(ISI+FFMC)^2+(temp+FFMC)^2, data = xtm_fire)
+summary(lm3)
+
+# compare models
+anova(lm1, lm2)
+anova(lm1, lm3)
+anova(lm2, lm3)
+
+#####################
+#
+# Practice Problems
+#
+####################
+
+# 1. Create a new subset that includes only with an ISI (inital spread index) greater than the median
+
+# 2. Create a linear model with area and wind as your predictors, and area as your response
+# Call this model lm1_ISI
+
+# 3. Determine if there are any correlated attributes
+
+# 4. If there are any correlated attributes add the interaction between them to a new model 
+# Call this model lm2_ISI
+
+# 5. Compare these two models and determine which model is better at predicting the size of the area burned