EDA of Red Wines quality by Oleksandr Brazhenko

# Load all of the packages that you end up using
# in your analysis in this code chunk.
library(ggplot2)
library(lsr)
library(GGally)
library(RColorBrewer)
library(grid)
library(gridExtra)
library(ellipse)
library(MASS)
library(ellipse)
library(lattice)
library(memisc)

## 
## Attaching package: 'memisc'

## The following objects are masked from 'package:stats':
## 
##     contrasts, contr.sum, contr.treatment

## The following object is masked from 'package:base':
## 
##     as.array

library(reshape2)
library(scales)

## 
## Attaching package: 'scales'

## The following object is masked from 'package:memisc':
## 
##     percent

#function to create hist object 
plot_hist_one_var <- function(x_str = "fixed.acidity", t, xl) {
  ggplot( data = rw, aes_string(x = x_str)) +
    geom_histogram(color="white", fill="#3366FF")+
    ggtitle(t)+
    xlab(xl)+
    theme(plot.title = element_text(size=12))
}

#function to create box plot object
box_plot_func <- function(x_str = "quality.rank", 
                          y_str, xl = "Quality rank",yl,t){
  ggplot( data = rw, aes_string(x=x_str, y = y_str)) +
    geom_boxplot()+
    xlab(xl)+
    ylab(yl)+ 
    ggtitle(t)
}

#function to create sactterplot object 
point_plot_func <- function( col = "quality.rank",
                            x_str, y_str, xl, yl, t, rvrs = T, 
                            lt = "Quality rank" ){
  ggplot( data = rw, aes_string(x=x_str, y = y_str, color=col)) +
    geom_point(size=3, alpha=0.5, position="jitter")+
    xlab(xl)+
    ylab(yl)+    
    ggtitle(t)+
    scale_color_brewer(palette="Spectral",
                       guide = guide_legend(title = lt, 
                                            reverse = rvrs))
}

#function to create density plot object 
density_plot_func <- function(d = rw,x_str,col,xl,t){
  ggplot( data = d, aes_string(x = x_str, color=col ))+
    geom_density(size = 1)+
    xlab(xl)+    
    ggtitle(t)  
}

Analysis

Main purpose of this data set analysis is figure out which chemical chracteristics have influence on red wine quality.

Explore the following parameters:

• fixed acidity (tartaric acid - g / dm^3)
• volatile acidity (acetic acid - g / dm^3)
• citric acid (g / dm^3)
• residual sugar (g / dm^3)
• chlorides (sodium chloride - g / dm^3
• free sulfur dioxide (mg / dm^3)
• total sulfur dioxide (mg / dm^3)
• density (g / cm^3)
• pH
• sulphates (potassium sulphate - g / dm3)
• alcohol (% by volume)

Output variable (based on sensory data): * quality (score between 0 and 10)

Load the Data

rw <- read.csv('/home/alex/ml/data/wineQualityReds.csv')

Summary of the Data Set

names(rw)

##  [1] "X"                    "fixed.acidity"        "volatile.acidity"    
##  [4] "citric.acid"          "residual.sugar"       "chlorides"           
##  [7] "free.sulfur.dioxide"  "total.sulfur.dioxide" "density"             
## [10] "pH"                   "sulphates"            "alcohol"             
## [13] "quality"

summary(rw)

##        X          fixed.acidity   volatile.acidity  citric.acid   
##  Min.   :   1.0   Min.   : 4.60   Min.   :0.1200   Min.   :0.000  
##  1st Qu.: 400.5   1st Qu.: 7.10   1st Qu.:0.3900   1st Qu.:0.090  
##  Median : 800.0   Median : 7.90   Median :0.5200   Median :0.260  
##  Mean   : 800.0   Mean   : 8.32   Mean   :0.5278   Mean   :0.271  
##  3rd Qu.:1199.5   3rd Qu.: 9.20   3rd Qu.:0.6400   3rd Qu.:0.420  
##  Max.   :1599.0   Max.   :15.90   Max.   :1.5800   Max.   :1.000  
##  residual.sugar     chlorides       free.sulfur.dioxide
##  Min.   : 0.900   Min.   :0.01200   Min.   : 1.00      
##  1st Qu.: 1.900   1st Qu.:0.07000   1st Qu.: 7.00      
##  Median : 2.200   Median :0.07900   Median :14.00      
##  Mean   : 2.539   Mean   :0.08747   Mean   :15.87      
##  3rd Qu.: 2.600   3rd Qu.:0.09000   3rd Qu.:21.00      
##  Max.   :15.500   Max.   :0.61100   Max.   :72.00      
##  total.sulfur.dioxide    density             pH          sulphates     
##  Min.   :  6.00       Min.   :0.9901   Min.   :2.740   Min.   :0.3300  
##  1st Qu.: 22.00       1st Qu.:0.9956   1st Qu.:3.210   1st Qu.:0.5500  
##  Median : 38.00       Median :0.9968   Median :3.310   Median :0.6200  
##  Mean   : 46.47       Mean   :0.9967   Mean   :3.311   Mean   :0.6581  
##  3rd Qu.: 62.00       3rd Qu.:0.9978   3rd Qu.:3.400   3rd Qu.:0.7300  
##  Max.   :289.00       Max.   :1.0037   Max.   :4.010   Max.   :2.0000  
##     alcohol         quality     
##  Min.   : 8.40   Min.   :3.000  
##  1st Qu.: 9.50   1st Qu.:5.000  
##  Median :10.20   Median :6.000  
##  Mean   :10.42   Mean   :5.636  
##  3rd Qu.:11.10   3rd Qu.:6.000  
##  Max.   :14.90   Max.   :8.000

class(rw$quality)

## [1] "integer"

Observations from the Summary

• Residual sugar maximum value is 15.5 g / dm^3. So, there are no sweet wines in this data set
• Huge dispersion of Total SO2 from 6 to 289.
• alcohol varies from 8.4 to 14.9 %
• pH range 2.74 - 4.0
• quality minimum value is 3 and maximum is 8

Understanding the individual variables in the data set

New categorical variable

I’d like to classify a quality variable into a larger bin, such as (4-5]; (5-6]; (7-8]. It makes analysis more deeper and specific. Lets create new categorical variable called rw$quality.rank with values “Low”, “Middle” and “High”.

rw$quality.rank <- cut(rw$quality, c(3,4,6,10), labels = c("low","middle","high"), include.lowest = T)

summary(rw$quality.rank)

##    low middle   high 
##     63   1319    217

#convert quality variable into categorical variable
#rw$quality <- integer(rw$quality)

Quality distribution

 ggplot( data = rw, aes(x = quality)) +
    geom_bar(color="white", fill="#3366FF")+
    ggtitle('Quality distribution')+
    xlab('Quality')+
    theme(plot.title = element_text(size=12))

Understanding the distribution of single variables

plot_fix_acid = plot_hist_one_var(t = "Fixed acidity distribution",
                                  x_str = "fixed.acidity",
                                  xl = "Fixed acidity (g / dm^3)")+
  coord_cartesian(xlim=c(quantile(rw$fixed.acidity, 0.01),
                         quantile(rw$fixed.acidity, 0.99)))

plot_vol_acid = plot_hist_one_var(x_str = "volatile.acidity",
                          t = "Volatile acidity distribution",
                          xl = "Volatile acidity (g / dm^3)") +
                     coord_cartesian(xlim=c(0,1))

plot_cit_acid = plot_hist_one_var(x_str = "citric.acid",
                          t = "Citric acidity distribution",
                          xl = "Citric acidity (g / dm^3)") +
                     coord_cartesian(xlim=c(0,0.75))

plot_res_sugar = plot_hist_one_var(x_str = "residual.sugar",
                          t = "Residual sugar distribution",
                          xl = "Residual sugar (g / dm^3)") +
                     scale_x_continuous(breaks = seq(1,5,0.5))+  
                     coord_cartesian(xlim=c(1,5))

plot_chlorides = plot_hist_one_var(x_str = "chlorides",
                          t = "Chlorides distribution",
                          xl = "Chlorides (g / dm^3)") +
                     coord_cartesian(xlim=c(0,0.2))

plot_so2 = plot_hist_one_var(x_str = "total.sulfur.dioxide",
                          t = "Total sulfur dioxide distribution",
                          xl = "Total sulfur dioxide (mg / dm^3)") +
                      coord_cartesian(xlim=c(0,150))

plot_ph = plot_hist_one_var(x_str = "pH",
                          t = "pH distribution",
                          xl = "pH") +
                      scale_x_continuous(breaks = seq(0,4,0.2))+
                      coord_cartesian(xlim=c(2.8,3.8))

plot_alcohol = plot_hist_one_var(x_str = "alcohol",
                          t = "Alcohol distributionn",
                          xl = "Alcohol (% by volume)")

grid.arrange(plot_fix_acid,plot_vol_acid,plot_cit_acid,plot_res_sugar,ncol=2)

## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

grid.arrange(plot_chlorides,plot_so2,plot_ph,plot_alcohol,ncol=2)

## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

Observations from distributions

• The most of variables have normal distribution
• The quality values seems to dispaly a normal distribution. Most observations have values equial to 5 or 6
• Citric acid has uniform distribution with big count values in 0
• volatile acidity have an values between 0.4 and 0.6
• most of wines consists residual sugar from 1.5 to 3.0 g/dm^3
• total SO2 mostly lie under the 50 g/dm^3 level
• pH also have strong normal distribution shape

Understanding of two variables relations

Correlation between variables

correlate(rw)

## 
## CORRELATIONS
## ============
## - correlation type:  pearson 
## - correlations shown only when both variables are numeric
## 
##                           X fixed.acidity volatile.acidity citric.acid
## X                         .        -0.268           -0.009      -0.154
## fixed.acidity        -0.268             .           -0.256       0.672
## volatile.acidity     -0.009        -0.256                .      -0.552
## citric.acid          -0.154         0.672           -0.552           .
## residual.sugar       -0.031         0.115            0.002       0.144
## chlorides            -0.120         0.094            0.061       0.204
## free.sulfur.dioxide   0.090        -0.154           -0.011      -0.061
## total.sulfur.dioxide -0.118        -0.113            0.076       0.036
## density              -0.368         0.668            0.022       0.365
## pH                    0.136        -0.683            0.235      -0.542
## sulphates            -0.125         0.183           -0.261       0.313
## alcohol               0.245        -0.062           -0.202       0.110
## quality               0.066         0.124           -0.391       0.226
## quality.rank              .             .                .           .
##                      residual.sugar chlorides free.sulfur.dioxide
## X                            -0.031    -0.120               0.090
## fixed.acidity                 0.115     0.094              -0.154
## volatile.acidity              0.002     0.061              -0.011
## citric.acid                   0.144     0.204              -0.061
## residual.sugar                    .     0.056               0.187
## chlorides                     0.056         .               0.006
## free.sulfur.dioxide           0.187     0.006                   .
## total.sulfur.dioxide          0.203     0.047               0.668
## density                       0.355     0.201              -0.022
## pH                           -0.086    -0.265               0.070
## sulphates                     0.006     0.371               0.052
## alcohol                       0.042    -0.221              -0.069
## quality                       0.014    -0.129              -0.051
## quality.rank                      .         .                   .
##                      total.sulfur.dioxide density     pH sulphates alcohol
## X                                  -0.118  -0.368  0.136    -0.125   0.245
## fixed.acidity                      -0.113   0.668 -0.683     0.183  -0.062
## volatile.acidity                    0.076   0.022  0.235    -0.261  -0.202
## citric.acid                         0.036   0.365 -0.542     0.313   0.110
## residual.sugar                      0.203   0.355 -0.086     0.006   0.042
## chlorides                           0.047   0.201 -0.265     0.371  -0.221
## free.sulfur.dioxide                 0.668  -0.022  0.070     0.052  -0.069
## total.sulfur.dioxide                    .   0.071 -0.066     0.043  -0.206
## density                             0.071       . -0.342     0.149  -0.496
## pH                                 -0.066  -0.342      .    -0.197   0.206
## sulphates                           0.043   0.149 -0.197         .   0.094
## alcohol                            -0.206  -0.496  0.206     0.094       .
## quality                            -0.185  -0.175 -0.058     0.251   0.476
## quality.rank                            .       .      .         .       .
##                      quality quality.rank
## X                      0.066            .
## fixed.acidity          0.124            .
## volatile.acidity      -0.391            .
## citric.acid            0.226            .
## residual.sugar         0.014            .
## chlorides             -0.129            .
## free.sulfur.dioxide   -0.051            .
## total.sulfur.dioxide  -0.185            .
## density               -0.175            .
## pH                    -0.058            .
## sulphates              0.251            .
## alcohol                0.476            .
## quality                    .            .
## quality.rank               .            .

Correlation plot

#craete new dataframe without categorical vars 
numeric_vars <- names(rw) %in% c("quality.rank", "quality")
rw_num <- rw[!numeric_vars]
ctab <- cor(rw_num)
#plot which show correlation in more convenient way
colorfun <- colorRamp(c("#CC0000","white","#3366CC"), space="Lab")
plotcorr(ctab, mar = c(0,0,0,0), col=rgb(colorfun((ctab+1)/2), 
                                         maxColorValue=255))

ggpairs function

#set.seed(278812)
#rw_samp <- rw[sample(1:length(rw$quality), 1000), ]
#ggpairs(rw_samp, axisLabels = "internal", 
#        params = c(shape = I('.'), outlier.shape = I('.')),
#        diag = list(prams = c(size = 1)))

Variables has the following correlations:

• fixed.acidity: positive correlation with citric.acid, density and negative correlation with pH
• volatile.acidity: negative correlation with citric.acid
• citric.acid: positive correlation with density and negative correlation with fixed.acidity, volatile.acidity, pH
• chlorides: positive correlation with sulphates
• pH: negative correlation with fixed.acidity, citric.acid, chlorides, density
• alcohol: negative correlation with density
• desity: positive correlation with fixed.acidity, citric.acid, residual.sugar and negative correlation with alcohol

Variables by quality ranking

Citric acid by quality rank

box_plot_func(y_str = "citric.acid", 
              yl = "Citric acid",
              t = "Citric acid by quality rank")

Quality wines have higher level of Citric acid. At the same time max value not bigger than 1.0 and 3rd quartile equal to 0.42. Outliers almost absent.

SO2 by quality rank

box_plot_func(y_str = "total.sulfur.dioxide", 
              yl = "Total SO2",
              t = "Total SO2 by quality rank")

No conclusions from Total SO2 distribution. Maybe, the reason is the values lie under the obvious limits in most of observations.

Residual sugar by quality rank

box_plot_func(y_str = "residual.sugar", 
              yl = "Residual sugar",
              t = "Residual sugar by quality rank")

Set limits and zoom the plot

box_plot_func(y_str = "residual.sugar", 
              yl = "Residual sugar",
              t = "Residual sugar by quality rank")+
  coord_cartesian(ylim=c(1.5,3))

by(rw$residual.sugar, rw$quality.rank,  summary)

## rw$quality.rank: low
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   1.200   1.900   2.100   2.685   2.950  12.900 
## -------------------------------------------------------- 
## rw$quality.rank: middle
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   0.900   1.900   2.200   2.504   2.600  15.500 
## -------------------------------------------------------- 
## rw$quality.rank: high
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   1.200   2.000   2.300   2.709   2.700   8.900

Residual sugar median have bigger values for quality wines. At the same time it has a huge dispersion, especially for poor wines. In this case 1st Qu. = 1.9 which lower than the “middle” and “high” rank of wines has. The 3rd Qu. = 2.95 which bigger than the other wines. Sugar max value is 15.5 so, there’s no sweet wines in this dataset (as we know, to be sweet wine should consist more than 45 g/m^3 sugar)

pH by quality rank

box_plot_func(y_str = "pH", 
              yl = "pH",
              t = "pH by quality rank")+
  coord_cartesian(ylim=c(3.1,3.6))

by(rw$pH, rw$quality.rank,  summary)

## rw$quality.rank: low
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   2.740   3.300   3.380   3.384   3.500   3.900 
## -------------------------------------------------------- 
## rw$quality.rank: middle
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   2.860   3.210   3.310   3.311   3.400   4.010 
## -------------------------------------------------------- 
## rw$quality.rank: high
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   2.880   3.200   3.270   3.289   3.380   3.780

pH has higher values for poor wines.

Alcohol by quality rank

box_plot_func(y_str = "alcohol", 
              yl = "Alcohol",
              t = "Alcohol by quality rank")+
    coord_cartesian(ylim=c(9,13))

by(rw$alcohol, rw$quality.rank,  summary)

## rw$quality.rank: low
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    8.40    9.60   10.00   10.22   11.00   13.10 
## -------------------------------------------------------- 
## rw$quality.rank: middle
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    8.40    9.50   10.00   10.25   10.90   14.90 
## -------------------------------------------------------- 
## rw$quality.rank: high
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    9.20   10.80   11.60   11.52   12.20   14.00

Significant difference of alcohol value between “high” quality wines and the others. At the same time, wine with “low” and “middle” quality rank have distribution and median pretty similar.

Sulphates by quality rank

box_plot_func(y_str = "sulphates", 
              yl = "Sulphates",
              t = "Sulphates by quality rank")

Set limits and zoom the plot

box_plot_func(y_str = "sulphates", 
              yl = "Sulphates",
              t = "Sulphates by quality rank")+
  coord_cartesian(ylim=c(0.4,0.9))

by(rw$sulphates, rw$quality.rank,  summary)

## rw$quality.rank: low
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##  0.3300  0.4950  0.5600  0.5922  0.6000  2.0000 
## -------------------------------------------------------- 
## rw$quality.rank: middle
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##  0.3700  0.5400  0.6100  0.6473  0.7000  1.9800 
## -------------------------------------------------------- 
## rw$quality.rank: high
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##  0.3900  0.6500  0.7400  0.7435  0.8200  1.3600

There are unexpected result a little bit. Quality rank depends on sulphates value. I’d assumed sulphates makes wine worst, but in this case it isn’t confirmed. Examine this value later.

Chlorides by quality rank

box_plot_func(y_str = "chlorides", 
              yl = "Chlorides",
              t = "Chlorides by quality rank")+ 
  coord_cartesian(ylim=c(0.05,0.1))

Chlorides have many outliers.

Observations from two variables exploration

Seems that wines quality depend on the following characteristics:

• alcohol;
• sulphates;
• Volatile acidity;
• Citric acid

Let’s examine these variables deeper.

Explore many variables

Examine pH more closely

Chlorides vs pH by Quality rank

point_plot_func( x_str = "pH", 
                 y_str = "chlorides", 
                 xl = "pH", 
                 yl = "Chorides", 
                 t = "Chlorides vs pH by Quality rank", 
                 rvrs = F)+
  coord_cartesian(ylim=c(0,0.2))

No conclusions from this plot

Sulphates vs pH by Quality rank

point_plot_func( x_str = "pH", 
                 y_str = "sulphates", 
                 xl = "pH", 
                 yl = "Sulphates", 
                 t = "Sulphates vs pH by Quality rank")

It looks like wines with higher sulphates value will be more quality with the equal level of pH. One more evidence of sulphates influence on quality.

Citric acid vs pH by Quality rank

point_plot_func( x_str = "pH", 
                 y_str = "citric.acid", 
                 xl = "pH", 
                 yl = "Citric acid", 
                 t = "Citric acid vs pH by Quality rank")

Most of HQ wines lie above 0.25 of Citric acid value

Volatile acidity vs pH by Quality rank

point_plot_func( x_str = "pH", 
                 y_str = "volatile.acidity", 
                 xl = "pH", 
                 yl = "Volatile acidity", 
                 t = "Volatile acidity vs pH by Quality rank",
                 rvrs = F)

Looks like wines with lower volatile acidity value will be of higher quality with the equal level of pH.

Examine sulphates more closely

Sulphates vs SO2 by Quality rank

point_plot_func( x_str = "total.sulfur.dioxide", 
                 y_str = "sulphates", 
                 xl = "Total SO2", 
                 yl = "Sulphates", 
                 t = "Sulphates vs Total SO2 by Quality rank")+
    coord_cartesian(ylim=c(0,1.5), xlim=c(0,100))

Another confirmation of sulphates influence. Seems like quality wines mostly consists no more than 50g of total SO2.

Sulphates vs volatile acidity by Quality rank

point_plot_func( x_str = "volatile.acidity", 
                 y_str = "sulphates", 
                 xl = "Volatile acidity", 
                 yl = "Sulphates", 
                 t = "Sulphates vs Volatile acidity by Quality rank")

Quality wines lies under the 0.4 volatile acidity. At the same time high quality wines has a big dispersion of sulphates than the other wines.

Sulphates vs alcohol by Quality rank

point_plot_func( x_str = "alcohol", 
                 y_str = "sulphates", 
                 xl = "Alcohol", 
                 yl = "Sulphates", 
                 t = "Sulphates vs Alcohol by Quality rank")

Another interesting plot. All quality wines placed in top right corner with higher level of sulfates and alcohol.

Examine alcohol more closely

Density vs alcohol by Quality rank

point_plot_func( x_str = "alcohol", 
                 y_str = "density", 
                 xl = "Alcohol", 
                 yl = "density", 
                 t = "Density vs alcohol by Quality rank")

Looks like density reduced in dependence on alcohol level growth

Density vs residual sugar by Quality rank

point_plot_func( x_str = "residual.sugar", 
                 y_str = "density", 
                 xl = "Residual sugar", 
                 yl = "Density", 
                 t = "Density vs Residual sugar Quality rank")+
  coord_cartesian(xlim=c(1,4))

No conclusions from this plot

Residual sugar vs Alcohol by Density

ggplot( data = rw, aes(x=alcohol, y = residual.sugar, color=density)) +
  geom_point(size=3, alpha=0.5, position="jitter")+
  xlab("Alcohol")+
  ylab("Residual sugar")+
  ggtitle("Residual sugar vs Alcohol by Density")+
  coord_cartesian(ylim=c(1,4))

No conclusions from this plot

Residual sugar vs Alcohol by Quality rank

point_plot_func( x_str = "alcohol", 
                 y_str = "residual.sugar", 
                 yl = "Residual sugar", 
                 xl = "Alcohol", 
                 t = "Residual sugar vs Alcohol by Quality rank")+
  coord_cartesian(ylim=c(1,4))

No conclusions from this plot

Citric acid vs alcohol by Quality rank

point_plot_func( x_str = "alcohol", 
                 y_str = "citric.acid", 
                 yl = "Citric acid", 
                 xl = "Alcohol", 
                 t = "Citric acid vs Alcohol by Quality rank")+
  coord_cartesian(ylim=c(0.05,0.15))

No conclusions from this plot

Volatile acidity vs alcohol by Quality rank

point_plot_func( x_str = "alcohol", 
                 y_str = "volatile.acidity", 
                 yl = "Volatile acidity", 
                 xl = "Alcohol", 
                 t = "Volatile acidity vs alcohol by Quality rank")

There are confirmed thesis wines more quality with lower volatile acidity level.

Additional analysis of high quality wines

#craete subset with only HQ rank wines
hqrw = subset(rw, quality.rank == "high")

ggplot()+
  geom_point(data = hqrw, aes(x=volatile.acidity, 
                              y = citric.acid),
             color = "#3366FF", 
             size=3)+
  ylab("Citric acid")+
  xlab("Volatle acidity")+    
  ggtitle("Citric acid vs Volatile acidity for high quality rank wines")+
  scale_x_continuous(breaks = seq(0,1.6,0.2))+  
  scale_y_continuous(breaks = seq(0,0.75,0.25))+  
  geom_rect(aes(xmin=0.4,xmax=Inf,ymin=0,ymax=0.25),alpha=0.2,fill="red")

This plot show mostly of HQ wines placed in zone with high level of citric acid (>0.25) and low level of volatile acidity (<0.4). However, there are some HQ wines which break this rule. They placed in red square on the plot.

I’d like to compare these two data sets and try to figure out which parameters are distinguished and which characteristics contribute to save high quality of wine even in case of high volatile acidity and low citric acid. Create new categorical variable in new data set. It’ll be defined whether wine lie in “red zone” or not. Let’s call it hqrw$is.outliers

hqrw$is.outliers <- ifelse(hqrw$citric.acid<0.25 & hqrw$volatile.acidity>0.4,
                        "y", "n")

Let’s look at our variables

density_plot_func(d = hqrw,
                  col = "is.outliers",
                  x_str = "pH",
                  xl = "pH",
                  t = "Density of pH for high quality rank wines")

by(hqrw$pH, hqrw$is.outliers,  summary)

## hqrw$is.outliers: n
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   2.880   3.170   3.240   3.241   3.320   3.780 
## -------------------------------------------------------- 
## hqrw$is.outliers: y
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   3.320   3.370   3.460   3.476   3.560   3.720

pH level significantly differ between these two groups. 1st Qu. of “red zones” values equal to 3.37. It’s higher than the 3rd Qu. (3.32) of the other part of data set. Before I noticed worse wines has higher pH value. In this case HQ wines have higher level of pH. So, it’s very interesting finding.

Let’s look on the other variables.

HQ_sulph_density = density_plot_func(d = hqrw,
                  col = "is.outliers",
                  x_str = "sulphates",
                  xl = "Sulphates",
                  t = "Sulphates density by HQ wine group")+
            theme(plot.title = element_text(size=10))  
  
HQ_totalso_density = density_plot_func(d = hqrw,
                  col = "is.outliers",
                  x_str = "total.sulfur.dioxide",
                  xl = "Total SO2",
                  t = "Total SO2 density by HQ wine group")+
            theme(plot.title = element_text(size=10))  

HQ_freeso_density = density_plot_func(d = hqrw,
                  col = "is.outliers",
                  x_str = "free.sulfur.dioxide",
                  xl = "Free SO2",
                  t = "Free SO2 density by HQ wine group")+
            theme(plot.title = element_text(size=10))  

HQ_chlorides_density = density_plot_func(d = hqrw,
                  col = "is.outliers",
                  x_str = "chlorides",
                  xl = "Chlorides",
                  t = "Chlorides density by HQ wine group")+
            theme(plot.title = element_text(size=10))  

grid.arrange(HQ_chlorides_density,
             HQ_freeso_density,
             HQ_totalso_density,
             HQ_sulph_density,
             ncol=2)

by(hqrw$total.sulfur.dioxide, hqrw$is.outliers,  summary)

## hqrw$is.outliers: n
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    7.00   16.00   25.00   33.31   42.00  289.00 
## -------------------------------------------------------- 
## hqrw$is.outliers: y
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   11.00   18.00   33.00   41.11   49.25  106.00

by(hqrw$free.sulfur.dioxide, hqrw$is.outliers,  summary)

## hqrw$is.outliers: n
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    3.00    6.00   10.00   13.73   18.00   54.00 
## -------------------------------------------------------- 
## hqrw$is.outliers: y
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    3.00   10.00   13.50   14.98   17.25   37.00

by(hqrw$chlorides, hqrw$is.outliers,  summary)

## hqrw$is.outliers: n
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
## 0.01200 0.06300 0.07400 0.07714 0.08500 0.35800 
## -------------------------------------------------------- 
## hqrw$is.outliers: y
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
## 0.04200 0.05400 0.06500 0.07109 0.08700 0.14300

Final Plots and Summary

Plot One

final_plot_one_1 = density_plot_func(col = "quality.rank",
                  x_str = "alcohol",
                  xl = "Alcohol",
                  t = "Alcohol density\n by quality rank")+
  theme(legend.justification=c(1,1), 
        legend.position=c(1,1))  

final_plot_one_2 = point_plot_func( x_str = "alcohol", 
                 y_str = "volatile.acidity", 
                 yl = "Volatile acidity", 
                 xl = "Alcohol", 
                 t = "Volatile acidity vs alcohol\n by quality rank",
                 rvrs = F)+
  theme(legend.justification=c(1,1), 
        legend.position=c(1,1)) 

grid.arrange(final_plot_one_1, final_plot_one_2, ncol=2)

HQ wines contain more alcohol than the other. However, we shouldn’t forget about acidity. In case of wines with alcohol more than 10% HQ wines placed in zone with low volatile acidity.

Plot Two

final_plot_two_1 = ggplot( data = rw, 
                           aes(x=quality.rank, 
                               y = sulphates, 
                               fill = quality.rank)) +
  geom_boxplot()+
  xlab("Quality rank")+
  ylab("Sulphates (g / dm^3)")+
  ggtitle("Sulphates by quality rank")+
  coord_cartesian(ylim=c(0.4,0.9))+
  theme(legend.justification=c(1,0), 
        legend.position=c(1,0))

final_plot_two_2 = point_plot_func( x_str = "pH", 
                 y_str = "sulphates", 
                 yl = "Sulphates", 
                 xl = "pH", 
                 t = "Sulphates vs pH by quality rank")+
  coord_cartesian(ylim=c(0.2,1.2))+ 
  theme(legend.justification=c(1,0), 
        legend.position=c(1,0))

grid.arrange(final_plot_two_1, final_plot_two_2, ncol=2)

Quality rank depends on sulphates value and this is unexpected result a little bit . If we look at sulphates vs pH level, we will find HQ wines contain more sulphates at the same level of pH.

Plot Three

final_plot_tree_1 = ggplot()+
  geom_point(alpha = 0.5, 
             data = rw, 
             aes(x=volatile.acidity,
                 y = citric.acid, 
                 color = quality.rank), 
             size=2)+
  scale_x_continuous(breaks = seq(0,1.6,0.2))+    
  ylab("Citric acid")+
  xlab("Volatle acidity")+    
  ggtitle("Citric acid vs Volatile acidity by quality rank")+
  scale_color_brewer(palette="Spectral",
                     type = 'div',
                     guide = guide_legend(title = 'Quality rank',
                                          reverse = T))+  
  geom_rect(aes(xmin=0.4,xmax=Inf,ymin=0,ymax=0.25),alpha=0.1,fill="red") +
  theme(plot.title = element_text(size=12),
        legend.justification=c(1,1), 
        legend.position=c(1,1))


final_plot_tree_2 = density_plot_func(d = hqrw,
                  col = "is.outliers",
                  x_str = "pH",
                  xl = "pH",
                  t = "Density of pH for HQ rank wines")+
  theme(plot.title = element_text(size=12), 
        legend.justification=c(1,1), 
        legend.position=c(1,1))

grid.arrange(final_plot_tree_1, final_plot_tree_2, ncol=2)

HQ wines mainly have low level of volatile acidity and citric acid level between 0.25 and 0.75. However, some samples of HQ wines has different parameters.

Additional research of HQ wines subset and its outliers show interesting dependencies. If wines have high level of quality even in “red zone”, then pH value significantly increase. To reduce influence of volatile acidity on quality need to increase quantity of antioxidants. As a result we receive higher value of pH for HQ wines with high level of acidity. Antioxidants may be such data set parameters as Chlorides, Total and Free SO2 and Sulphates.

Summary

Three final plots demonstrate dependance wines quality from the several characteristics. As a result, the same quality wine level may have different proportions of SO2, sulphates, pH, acidity and alcohol.

Main findings and trends of red wine data set:

• high quality wines have higher level of alcohol;
• at the same time Volatile acidity level shouldn’t be higher than 0.4. In the case of high volatile acidity level need to raise pH value;
• most of HQ wines consist from 0.25 to 0.75 citric acid;
• sulphates level make wines more quality. However, they should lie under the 1.0 value;
• HQ wines don’t consist Total SO2 more than 50 g/dm^3.

Another interesting thing is poor wines has a high pH level, but in the third plot we saw opposite evidence. I think pH level makes balance to provide good taste for wine.

For example, we have pretty similar values of pH for the different wine quality. in main data set for poor wines: Min. 1st Qu. Median Mean 3rd Qu. Max. 2.740 3.300 3.380 3.384 3.500 3.900

in HQ wines subset for “outliers”: Min. 1st Qu. Median Mean 3rd Qu. Max. 3.320 3.370 3.460 3.476 3.560 3.720

So, we can’t examine wine quality based only on the single parameter.

Reflection

It was interesting to explore this data set. I’m not expert of wines. I even don’t drink. So, it was fascinatingly to determine which characteristics make wines good or poor without tasting :)

When I used different plot types I received more complex information about the same variable. (e.g. boxplot show different details rather than histogram or density plot). I never used boxplots before. So, it’s very useful tool to describe variable.

Exploring subset of HQ wines and outliers in the end of data analysis discovered some interesting dependencies. After that, I realized how parameters contribute to each other.

During analysis I faced such difficulties as:

• some R functions. I used manual, blogs etc. For example, to set parameters of ggpairs function;
• I don’t fully understood pH role in data set.

Moreover, there are several tricky moment in the data set:

• presence of input data errors (says in data description);
• a few observations of poor and high quality wines;
• all wines produced in the same region of Portugal. Thus, wines from France or the US may have other character istic values.

In this data set I didn’t find such features like age of wine, grape, soil acidity etc. Feasibly, one of them could show interesting correlations.

Possible future researches:

• go deeper in data set and explore such features as residual sugar, fixed acidity;
• continue to explore outliers and examine poor quality wines;
• apply different methods of regression analysis.

EDA really exciting and may take a huge time to research. You should to know time to stop.