Modeling2.Rmd

Jared
```{R}
if(!require(mlbench)){install.packages("mlbench"); require(mlbench)} # common datasets to use
if(!require(caret)){install.packages("caret", dependencies = c("Depends", "Suggests")); require(caret)} # ML package and its dependencies. This will take awhile!
if(!require(dplyr)){install.packages("dplyr"); require(dplyr)}
if(!require(ModelMetrics)){install.packages("ModelMetrics"); require(ModelMetrics)}
if(!require(recipes)){install.packages("recipes"); require(recipes)}
if(!require(DEoptimR)){install.packages("DEoptimR"); require(DEoptimR)}
set.seed(370)


```

```{r}
#Read in data
baseline <- read.csv(file="/Users/jaredpraino/Documents/University of Washington/Senior Year/INFO370/final_project/project-fragile-families/data/Baseline_Features.csv",header=TRUE,sep=",")
year1 <- read.csv(file="/Users/jaredpraino/Documents/University of Washington/Senior Year/INFO370/final_project/project-fragile-families/data/Year1_Features.csv", header=TRUE,sep=",")
year3 <- read.csv(file="/Users/jaredpraino/Documents/University of Washington/Senior Year/INFO370/final_project/project-fragile-families/data/Year3_Features.csv",header=TRUE,sep=",")
year5 <- read.csv(file="/Users/jaredpraino/Documents/University of Washington/Senior Year/INFO370/final_project/project-fragile-families/data/Year5_Features.csv",header=TRUE,sep=",")
year9 <- read.csv(file="/Users/jaredpraino/Documents/University of Washington/Senior Year/INFO370/final_project/project-fragile-families/data/Year9_Features.csv",header=TRUE,sep=",")
```

```{r}
#Select top 56 features and remove missing values
temp_parent_involvement <- year9 %>% select(idnum,t5c13a,t5c13b,t5c13c) 

#View(year9)

combine0 <- baseline %>% select(idnum, m1a4, m1a11a, m1a11b) %>% 
  filter(idnum>-1, m1a4>-1, m1a11a>-1, m1a11b>-1)

combine1 <- year1 %>% select(idnum,m2b18a,m2b18b,m2b18c,m2b18d,m2b18e,m2b18f,m2b18g,m2b18h,m2c3a,m2c3b, m2c3c,m2c3d,m2c3e,m2c3f,m2c3h,m2c3i,m2c3j) %>%    filter(m2b18a>-1,m2b18b>-1,m2b18c>-1,m2b18d>-1,m2b18e>-1,m2b18f>-1,m2b18g>-1,m2b18h>-1,m2c3a>-1,m2c3b>-1, m2c3c>-1,m2c3d>-1,m2c3e>-1,m2c3f>-1,m2c3h>-1,m2c3i>-1,m2c3j>-1)

combine3 <- year3 %>% select(idnum,m3b4a,m3b4b,m3b4c,m3b4d,m3b4e,m3b4f,m3b4g,m3b4h,m3b4i,m3b4k,m3b4l,m3b5,m3c3a,m3c3b,m3c3c,m3c3d,m3c3e,m3c3f,m3c3g,m3c3h,m3c3i,m3c3j,m3c3k,m3c3l) %>% 
  filter(m3b4a>-1,m3b4b>-1,m3b4c>-1,m3b4d>-1,m3b4e>-1,m3b4f>-1,m3b4g>-1,m3b4h>-1,m3b4i>-1,m3b4k>-1,m3b4l>-1,m3b5>-1,m3c3a>-1,m3c3b>-1,m3c3c>-1,m3c3d>-1,m3c3e>-1,m3c3f>-1,m3c3g>-1,m3c3h>-1,m3c3i>-1,m3c3j>-1,m3c3k>-1,m3c3l>-1)

combine5 <- year5 %>% select(idnum,m4b4a1,m4b4a2,m4b4a3,m4b4a4,m4b4a5,m4b4a6,m4b4a7,m4b4a8) %>% 
  filter(m4b4a1>-1,m4b4a2>-1,m4b4a3>-1,m4b4a4>-1,m4b4a5>-1,m4b4a6>-1,m4b4a7>-1,m4b4a8>-1)

combine9 <- year9 %>% select(idnum,t5c13a,t5c13b,t5c13c) %>% 
  filter(t5c13a>-1,t5c13b>-1,t5c13c>-1)

join1 <- inner_join(combine0,combine1, by="idnum")
join2 <- inner_join(join1,combine3, by="idnum")
join3 <- inner_join(join2,combine5, by="idnum")
finalCombined <- inner_join(join3,combine9, by="idnum")

# prepare new columns
# total activities
parent_involvement <- finalCombined %>% mutate(tally = 0)

# total grade
parent_involvement <- parent_involvement %>% mutate(score = t5c13a + t5c13b + t5c13c)


# add outcome variables
parent_involvement$t5c13a <- as.factor(parent_involvement$t5c13a)
parent_involvement$t5c13b <- as.factor(parent_involvement$t5c13b)
parent_involvement$t5c13c <- as.factor(parent_involvement$t5c13c)


#Subsets with grades from year 5 and year 9 isolated
finalSet <- read.csv(file="/Users/grant/Documents/INFO370/fragile_families/project-fragile-families/data/FeaturesWithYear9Year5MartialStatus.csv")


library(dplyr)
year9grades <- filter(finalSet, t5c13a>-1, t5c13b>-1, t5c13c>-1)
year5grades <- filter(finalSet, kind_a5>-1, kind_a6>-1, kind_a7>-1)
```


```{r}
####    Version 1:  Popularity of activites from Yuxing's feature selection

# ##  SET UP
# top  DPW - hand picked features in days/per/week format
#need more features
features <- c('m4b4a2','m3c3l','m4b4a8','m3b4l','m3b4a','m2c3j','m2c3f','m3c3g','m2c3e','m3b4d','m2c3a')

#

# top  YN - hand picked features in yes no format
#
# 

### View(parent_involvement)

# ##  FUNCTIONS

# COMPUTED MEAN
computed_mean <- function(column) {
  
  response <- parent_involvement[[column]]
  weighted_mean <- (sum(response==0)*0 + sum(response==1)*1 + sum(response==2)*2 + sum(response==3)*3 + sum(response==4)*4 +       sum(response==5)*5 + sum(response==6)*6 +sum(response==7)*7)/length(response)
  
  fltr <- response > weighted_mean
  parent_involvement[fltr,"tally"] <- parent_involvement[fltr,"tally"] + 1
  return(parent_involvement)
}


# ##  COMPUTATIONS
# Number of activities
weighted_means <- c(1:length(features))
for(i in 1:length(features)){
  column_name = features[i]
  parent_involvement <- computed_mean(column_name)
}


# Percentage of activities
length <- length(features)
parent_involvement <- parent_involvement %>% mutate(percentage = tally/11.0)
```


```{r}
#synthesize marital status to M (married) and D (divorced). This finished set can be found under data/FeaturesWithYear9Year5MartialStatus.csv"
marriageCol5 <- ifelse(year5grades$m4a4 == 1, "Married",
                ifelse(year5grades$m4a4 == 2, "Married",
                ifelse(year5grades$m4a4 == 3, "Divorced",
                ifelse(year5grades$m4a4 == 4, "Divorced",
                ifelse(year5grades$m4a4 == 5, "Divorced",
                ifelse(year5grades$m4a4 == 6, "Divorced", NA))))))

marriageCol9 <- ifelse(year9grades$m5a4 == 1, "Married",
                ifelse(year9grades$m5a4 == 2, "Married",
                ifelse(year9grades$m5a4 == 3, "Divorced",
                ifelse(year9grades$m5a4 == 4, "Divorced",
                ifelse(year9grades$m5a4 == 5, "Divorced",
                ifelse(year9grades$m5a4 == 6, "Divorced", NA))))))


year5grades$marrystat <- marriageCol5

year9grades$marrystat <- marriageCol9

#Change school performance to a binary value, 0 for below average, 1 for above average

year9gradesHighLow <- year9grades

year9gradesHighLow <- year9gradesHighLow %>% filter(t5c13c != 3, t5c13a != 3, t5c13b != 3)

year9gradesHighLow$abovebelowMath <- ifelse(year9gradesHighLow$t5c13c == 1, 0,
                                      ifelse(year9gradesHighLow$t5c13c == 2, 0,
                                      ifelse(year9gradesHighLow$t5c13c == 4, 1,
                                      ifelse(year9gradesHighLow$t5c13c == 5, 1,
                                             -1
                                      ))))
year9gradesHighLow$abovebelowReading <- ifelse(year9gradesHighLow$t5c13a == 1, 0,
                                      ifelse(year9gradesHighLow$t5c13a == 2, 0,
                                      ifelse(year9gradesHighLow$t5c13a == 4, 1,
                                      ifelse(year9gradesHighLow$t5c13a == 5, 1,
                                             -1
                                      ))))
year9gradesHighLow$abovebelowScience <- ifelse(year9gradesHighLow$t5c13b == 1, 0,
                                      ifelse(year9gradesHighLow$t5c13b == 2, 0,
                                      ifelse(year9gradesHighLow$t5c13b == 4, 1,
                                      ifelse(year9gradesHighLow$t5c13b == 5, 1,
                                             -1
                                             ))))
year9gradesHighLow$abovebelowScore <- ifelse(year9gradesHighLow$score == 1, 0,
                                      ifelse(year9gradesHighLow$score == 2, 0,
                                      ifelse(year9gradesHighLow$score == 3, 0,
                                      ifelse(year9gradesHighLow$score == 4, 0,
                                      ifelse(year9gradesHighLow$score == 5, 0,
                                      ifelse(year9gradesHighLow$score == 6, 0,
                                      ifelse(year9gradesHighLow$score == 7, 0,
                                      ifelse(year9gradesHighLow$score == 9, 1,
                                      ifelse(year9gradesHighLow$score == 10, 1,
                                      ifelse(year9gradesHighLow$score == 11, 1,
                                      ifelse(year9gradesHighLow$score == 12, 1,
                                      ifelse(year9gradesHighLow$score == 13, 1,
                                      ifelse(year9gradesHighLow$score == 14, 1,
                                      ifelse(year9gradesHighLow$score == 15, 1, NA))))))))))))))

#remove NA values
year9gradesHighLow <- year9gradesHighLow[!is.na(year9gradesHighLow$score), ]
```


```{r}
####    Verison 2: Popularity of all activities in the days/week format


# total grade


# Our predictor variables
features2 <- c('f1g9a',	'f1g9b',	'f1g9c',	'f1g9d',	'f1g9e',	'f1g9f',	'f1g9g',	'f1g9h',	'f1g9i',	'f1g9j',	'f1g9k',	'f1g9l','m2b18a',	'm2b18b',	'm2b18c',	'm2b18d',	'm2b18e',	'm2b18f',	'm2b18g',	'm2b18h','m2c3a','m3b4a',	'm3b4b',	'm3b4c',	'm3b4d',	'm3b4e',	'm3b4f',	'm3b4g',	'm3b4h',	'm3b4i',	'm3b4j',	'm3b4k',	'm3b4l',	'm3b4m','m3c3a','m3c3b',	'm3c3c',	'm3c3d',	'm3c3e',	'm3c3f',	'm3c3g',	'm3c3h',	'm3c3i',	'm3c3j',	'm3c3k',	'm3c3l',	'm3c3m','f3c3a',	'f3c3b',	'f3c3c',	'f3c3d',	'f3c3e',	'f3c3f',	'f3c3g',	'f3c3h',	'f3c3i',	'f3c3j',	'f3c3k',	'f3c3l',	'f3c3m','f3b4a',	'f3b4b',	'f3b4c',	'f3b4d',	'f3b4e',	'f3b4f',	'f3b4g',	'f3b4h',	'f3b4i',	'f3b4j',	'f3b4k',	'f3b4l',	'f3b4m','m4c3a',	'm4c3b',	'm4c3c',	'm4c3d',	'm4c3e',	'm4c3f','m4c3g',	'm4c3h','f4b4a1',	'f4b4a2',	'f4b4a3',	'f4b4a4',	'f4b4a5',	'f4b4a6',	'f4b4a7',	'f4b4a8',	'f4c3a',	'f4c3b',	'f4c3c',	'f4c3d',	'f4c3e',	'f4c3f',	'f4c3g',	'f4c3h','f2b36a',	'f2b36b',	'f2b36c',	'f2b36d',	'f2b36e',	'f2b36f',	'f2b36g',	'f2b36h',	'f2c3a',	'f2c3b',	'f2c3c',	'f2c3d',	'f2c3e',	'f2c3f',	'f2c3g',	'f2c3h',	'f2c3i',	'f2c3j','m2c3b',	'm2c3c',	'm2c3d',	'm2c3e',	'm2c3f',	'm2c3g',	'm2c3h',	'm2c3i',	'm2c3j')


parent_involvement2 <- read.csv("/Users/grant/Documents/INFO370/fragile_families/project-fragile-families/data/FeaturesSelected.csv")

## convert 200's values
# 201 => 7
# 202 => 3
# 203 => 2
# 204 => 1
# 205 => 0
parent_involvement2[parent_involvement2==205] <- 0
parent_involvement2[parent_involvement2==204] <- 1
parent_involvement2[parent_involvement2==203] <- 2
parent_involvement2[parent_involvement2==202] <- 3
parent_involvement2[parent_involvement2==201] <- 7
# remove negative values
parent_involvement2[parent_involvement2<0] <- NA


# grades and total grade
combine9_2 <- year9 %>% select(idnum,t5c13a,t5c13b,t5c13c) %>% 
  filter(t5c13a>-1,t5c13b>-1,t5c13c>-1)
combine9_2 <- combine9 %>% mutate(score=t5c13a + t5c13b + t5c13c)
#temp_save_df <- parent_involvement2

combine9_2$t5c13a <- as.factor(combine9_2$t5c13a)
combine9_2$t5c13b <- as.factor(combine9_2$t5c13b)
combine9_2$t5c13c <- as.factor(combine9_2$t5c13c)
combine9_2$score <- as.factor(combine9_2$score)

# merge
typeof(parent_involvement2$idnum)
combine9_2$idnum <- as.numeric(combine9_2$idnum)
parent_involvement2 <- inner_join(combine9_2, parent_involvement2, by="idnum")
parent_involvement2 <- parent_involvement2 %>% mutate(tally = 0)


#temp_save_df <- inner_join(combine9_2, parent_involvement2, by="idnum")
#temp_save_df_2 <- inner_join(combine9_2, parent_involvement2, by="idnum")
#temp_save_df <- temp_save_df %>% mutate(tally = 0)
#temp_save_df_2 <- temp_save_df_2 %>% mutate(tally = 0)

View(parent_involvement2)

# function for determining effort to tally activity and then tallying.
computed_mean2 <- function(column) {
   
  response <- na.omit(parent_involvement2[[column_name]])
  weighted_mean <- (sum(response==0)*0 + sum(response==1)*1 + sum(response==2)*2 + sum(response==3)*3 + sum(response==4)*4 +       sum(response==5)*5 + sum(response==6)*6 +sum(response==7)*7)/length(response)
  
  fltr <- response > weighted_mean
  parent_involvement2[fltr,"tally"] <- parent_involvement2[fltr,"tally"] + 1
  return(parent_involvement2)
}

# Add tallies for doing activities
for(i in 1:length(features2)){
  column_name = features2[i]
  parent_involvement2 <- computed_mean2(column_name)
}
# check data
View(parent_involvement2)


parent_involvement2 <- parent_involvement2 %>% mutate(percentage = tally/124.0)

#parent_involvement2 <- as.numeric(parent_involvement2$tally)
#parent_involvement2 <- as.numeric(parent_involvement2$t5c13c)



```

``` {r}
#Decision tree and confusion matrix

if(!require(rpart.plot)){install.packages("rpart.plot"); require(rpart.plot)} # only need this is dependencies of caret were not installed
ctrl <- trainControl(method = 'cv', number=5)
model_dec_tree <- train(t5c13c ~ ., 
                        data = parent_involvement2,
                        method = "rpart",
                        trControl = ctrl,
                        tuneLength = 10) # try 10 reasonable parameter values. No more guessing parameter ranges!

model_dec_tree

# looking at decision tree
prp(parent_involvement2$t5c13c)

# predictions using model
predict_iris <- predict(model_dec_tree, newdata = iris_test_x)

# performance (via confusion matrix)
confusionMatrix(predict_iris, iris_test_y$Species)
```



```{r}
#Clean year 5 data

finalSet <- read.csv("/Users/jaredpraino/Documents/University of Washington/Senior Year/INFO370/final_project/project-fragile-families/data/FeaturesWithYear9Year5MartialStatus.csv")

features_year_5 <- c('m4b4a2','m3c3l','m4b4a8','m3b4l','m3b4a','m2c3j','m2c3f','m3c3g','m2c3e','m3b4d','m2c3a')

year9grades <- filter(finalSet, t5c13a>-1, t5c13b>-1,t5c13c>-1)
year5grades <- filter(finalSet, kind_a5>-1, kind_a6>-1, kind_a7>-1)

year5grades <- year5grades %>% mutate(tally = 0)
# total grade
year5grades <- year5grades %>% mutate(scoreKind = t5c13a + t5c13b + t5c13c)
# add outcome variables
year5grades$t5c13a <- as.factor(year5grades$t5c13a)
year5grades$t5c13b <- as.factor(year5grades$t5c13b)
year5grades$t5c13c <- as.factor(year5grades$t5c13c)

year5grades[year5grades==205] <- 0
year5grades[year5grades==204] <- 1
year5grades[year5grades==203] <- 2
year5grades[year5grades==202] <- 3
year5grades[year5grades==201] <- 7
# remove negative values
year5grades[year5grades<0] <- NA


computed_mean_year_5 <- function(column) {
   
  response <- na.omit(year5grades[[column_name]])
  weighted_mean <- (sum(response==0)*0 + sum(response==1)*1 + sum(response==2)*2 + sum(response==3)*3 + sum(response==4)*4 +       sum(response==5)*5 + sum(response==6)*6 +sum(response==7)*7)/length(response)
  
  fltr <- response > weighted_mean
  year5grades[fltr,"tally"] <- year5grades[fltr,"tally"] + 1
  return(year5grades)
}


# Add tallies for doing activities
for(i in 1:length(features_year_5)){
  column_name = features_year_5[i]
  year5grades <- computed_mean_year_5(column_name)
}

#graph year 5 scores vs tallies of activities
ggplot(data = year5grades, aes(x = scoreKind, y = tally)) + geom_point() + geom_jitter(width = 0.2) +
  geom_smooth(method = "lm", color="black", aes(group=1))

cor(year5grades$score,year5grades$tally)

```

```{r}
#chi squared values
marriedYear9 <- filter(year9gradesHighLow, marrystat == "M")

divorcedYear9 <- filter(year9gradesHighLow, marrystat == "D")


count(marriedYear9, var = abovebelowScore)

count(divorcedYear9, var = abovebelowScore)


belowMarryRatio <- 166/687

aboveMarryRatio <- 521/687

belowDivRation <- 270/680

aboveDivRatio <- 410/680

paste("Below Marry Ratio: "  ,belowMarryRatio)
paste("Above Marry Ratio: "  , aboveMarryRatio)

paste("Below Divorce Ratio: " , belowDivRation)
paste("Above Divorce Ratio: " , aboveDivRatio)
```