ger.Rmd

---
title: "ger"
author: "Ernesto Carrella"
date: "February 16, 2019"
output: html_document
editor_options: 
  chunk_output_type: console
---

```{r setup, include=FALSE}
set.seed(0)
knitr::opts_chunk$set(echo = FALSE, warning= FALSE, message=FALSE,
                      cache=TRUE,cache.extra = knitr::rand_seed,
                      dpi = 300, fig.width = 13.3, fig.height = 10)
library(abctools)
source('./setup.R')
library(tidyverse)
bam_loess<-FALSE
FILTER_TO_1250<-FALSE
FILENAME<-"ger"
if(FILTER_TO_1250)
  FILENAME<-paste0(FILENAME,"_small")


```




```{r dataprep}
# I add to add a "row.number" column name to the file to read this correctly
dr_sumstat<-read_csv("./data/ger/ger.csv",
                     col_names=TRUE) %>% select_if(function(x){var(x)>0}) %>% select_if(function(x){length(unique(x))>20})




colnames(dr_sumstat) <-
  colnames(dr_sumstat) %>% make.names(unique=TRUE,allow_ = FALSE)


# I add to add a "row.number" column name to the file to read this correctly
dr_parameters<-
  read_csv("./data/ger/params_ger.csv",
                     col_names=TRUE) %>%
  dplyr::select(-`random-seed`)

colnames(dr_parameters) <-
  colnames(dr_parameters) %>% make.names(unique=TRUE,allow_ = FALSE)




total_data<-
  cbind(dr_parameters,dr_sumstat) 

if(FILTER_TO_1250){
  total_data<- total_data %>%
  filter(row_number()<=1250)
}

total_data<-
  total_data 



param_names<-colnames(dr_parameters)
x_names<-colnames(dr_sumstat)
```

# ABC

## Rejection


```{r rejection}
rejection_cross<-
  cross_validate_abc(total_data,ngroup=5,method="rejection",parameter_colnames = param_names)
print_table(rejection_cross)

```

## Rejection - semiauto


```{r rejection-sabc}
rejection_cross_sabc<-
  cross_validate_abc(total_data,ngroup=5,method="rejection",parameter_colnames = param_names,
                     semiauto = TRUE)
print_table(rejection_cross_sabc)

```



## Neural-Network

```{r neuralnet, eval=TRUE, message= FALSE, results="hide"}
neuralnet<-
  cross_validate_abc(total_data,ngroup=5,method="neuralnet",
                     hcorr=TRUE,parameter_colnames = param_names)

```
```{r, eval=TRUE}
print_table(neuralnet)

```

# Linear regression


## Degree 1

```{r lm}
lm_cross<-
  cross_validate_lm(total_data,ngroup=5,
                    parameter_colnames = param_names,x_names=x_names,
                     degree=1)
print_table(lm_cross)

```

# Random Forest



```{r rf}
rf_cross<-
  cross_validate_rf(total_data,ngroup=5,
                    parameter_colnames = param_names,x_names=x_names)
print_table(rf_cross)

```


# Random Forest (boot)



```{r rfboot, results="hide"}
rfboot_cross<-
  cross_validate_rfboot(total_data,ngroup=5,
                    parameter_colnames = param_names,x_names=x_names)

```

```{r rfboottable}
print_table(rfboot_cross)
```



# Loess Fit



```{r loess}
loess_cross<- cross_validate_loess(total_data,ngroup=5,
                    parameter_colnames = param_names,
                    x_names=x_names,
                    bam=bam_loess)
print_table(loess_cross)

```



## Rejection - 1D


```{r rejection-sabc1d}
rejection_cross_sabc1d<-
  cross_validate_abc(total_data,ngroup=5,method="rejection",parameter_colnames = param_names,
                     semiauto = TRUE,
                     satr=list(function(x){outer(x,Y=1,"^")}))
print_table(rejection_cross_sabc1d)

```


## Loclin


```{r loclin, eval=TRUE}
loclin_cross<-
  cross_validate_abc(total_data,ngroup=5,method="loclinear",
                     hcorr=TRUE,parameter_colnames = param_names,
                     semiauto = FALSE)
print_table(loclin_cross)

```

# Average



```{r average}
average_cross<- cross_validate_average(total_data,ngroup=5,
                    parameter_colnames = param_names,
                    x_names=x_names)
print_table(average_cross)

```

```{r, writetocsv, eval=TRUE}

crosses<-
  list(
    rejection = rejection_cross,
    sabc_4d = rejection_cross_sabc,
    sabc_1d = rejection_cross_sabc1d,
    loclin = loclin_cross, 
    neuralnet = neuralnet,
    lm = lm_cross,
    rf = rf_cross,
    loess = loess_cross,
    rfboot = rfboot_cross,
    average = average_cross
  )

write_csv(
  map2_df(crosses,names(crosses),~ as_data_frame(t(.x$error)) %>% 
            mutate(method=.y)) %>% gather(parameter,value,-method), 
  paste("./errors/",FILENAME,".csv",sep="")
)

write_csv(
  map2_df(crosses,names(crosses),~ as_data_frame(t(.x$contained)) %>% 
            mutate(method=.y)) %>% gather(parameter,value,-method),   paste("./contained/",FILENAME,".csv",sep="")
)

write_csv(
  map2_df(crosses,names(crosses),~ as_data_frame(t(.x$interval)) %>% 
            mutate(method=.y)) %>% gather(parameter,value,-method), 
  paste("./intervals/",FILENAME,".csv",sep="")
)
```