PyData-EDA/Investigating Factors Affecting Red Wine Quality/Red_Wine_Quality.rmd

What Makes A Good Wine?
========================================================

In this project, a data set of red wine quality is explored based on its 
physicochemical properties. The objective is to find physicochemical properties
that distinguish good quality wine from lower quality ones. An attempt to build
linear model on wine quality is also shown.

### Dataset Description
This tidy dataset contains 1,599 red wines with 11 variables on the chemical 
properties of the wine. Another variable attributing to the quality of wine is 
added; at least 3 wine experts did this rating. The preparation of the dataset 
has been described in [this link](https://goo.gl/HVxAzY).

```{r global_options, include=FALSE}
knitr::opts_chunk$set(fig.path='Figs/',
                      echo=FALSE, warning=FALSE, message=FALSE)
```

```{r echo=FALSE, message=FALSE, warning=FALSE, packages}
library(ggplot2)
library(gridExtra)
library(GGally)
library(ggthemes)
library(dplyr)
library(memisc)
```

First, the structure of the dataset is explored using ``summary`` and ``str`` 
functions.  
```{r echo=FALSE, warning=FALSE, message=FALSE, Load_the_Data}
wine <- read.csv("wineQualityReds.csv")
str(wine)
summary(wine)

# Setting the theme for plotting.
# theme_set(theme_minimal(10))

# Converting 'quality' to ordered type.
wine$quality <- ordered(wine$quality, 
                        levels=c(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10))
# Adding 'total.acidity'.
wine$total.acidity <- wine$fixed.acidity + wine$volatile.acidity
```

**The following observations are made/confirmed:** 

1. There are 1599 samples of Red Wine properties and quality values.

2. No wine achieves either a terrible (0) or perfect (10) quality score.

3. Citric Acid had a minimum of 0.0. No other property values were precisely 0.

4. Residual Sugar measurement has a maximum that is nearly 20 times farther
away from the 3rd quartile than the 3rd quartile is from the 1st. There is
a chance of a largely skewed data or that the data has some outliers.

5. The 'quality' attribute is originally considered an integer;
I have converted this field into an ordered factor which is much more
a representative of the variable itself.

6. There are two attributes related to 'acidity' of wine i.e. 'fixed.acidity' 
and 'volatile.acidity'. Hence, a combined acidity variable is added 
using ``data$total.acidity <- data$fixed.acidity + data$volatile.acidity``.

## Univariate Plots Section
To lead the univariate analysis, I’ve chosen to build a grid of histograms. 
These histograms represent the distributions of each variable in the dataset.

```{r echo=FALSE, warning=FALSE, message=FALSE, Univariate_Grid_Plot}
g_base <- ggplot(
  data = wine,
  aes(color=I('black'), fill=I('#990000'))
) 

g1 <- g_base +
  geom_histogram(aes(x = fixed.acidity), binwidth = 0.25) + 
  scale_x_continuous(breaks = seq(4, 16, 2)) + 
  coord_cartesian(xlim = c(4, 16))

g2 <- g_base +
  geom_histogram(aes(x = volatile.acidity), binwidth = 0.05) + 
  scale_x_continuous(breaks = seq(0, 2, 0.5)) +
  coord_cartesian(xlim = c(0, 2))

g3 <- g_base +
  geom_histogram(aes(x = total.acidity), binwidth = 0.25) + 
  scale_x_continuous(breaks = seq(0, 18, 1)) +
  coord_cartesian(xlim = c(4, 18))

g4 <- g_base +
  geom_histogram(aes(x = citric.acid), binwidth = 0.05) + 
  scale_x_continuous(breaks = seq(0, 1, 0.2)) +
  coord_cartesian(xlim = c(0, 1))

g5 <- g_base +
  geom_histogram(aes(x = residual.sugar), binwidth = 0.5) + 
  scale_x_continuous(breaks = seq(0, 16, 2)) +
  coord_cartesian(xlim = c(0, 16))

g6 <- g_base +
  geom_histogram(aes(x = chlorides), binwidth = 0.01) + 
  scale_x_continuous(breaks = seq(0, 0.75, 0.25)) +
  coord_cartesian(xlim = c(0, 0.75))

g7 <- g_base +
  geom_histogram(aes(x = free.sulfur.dioxide), binwidth = 2.5) + 
  scale_x_continuous(breaks = seq(0, 75, 25)) +
  coord_cartesian(xlim = c(0, 75))

g8 <- g_base +
  geom_histogram(aes(x = total.sulfur.dioxide), binwidth = 10) + 
  scale_x_continuous(breaks = seq(0, 300, 100)) +
  coord_cartesian(xlim = c(0, 295))

g9 <- g_base +
  geom_histogram(aes(x = density), binwidth = 0.0005) + 
  scale_x_continuous(breaks = seq(0.99, 1.005, 0.005)) +
  coord_cartesian(xlim = c(0.99, 1.005))

g10 <- g_base +
  geom_histogram(aes(x = pH), binwidth = 0.05) + 
  scale_x_continuous(breaks = seq(2.5, 4.5, 0.5)) +
  coord_cartesian(xlim = c(2.5, 4.5))

g11 <- g_base +
  geom_histogram(aes(x = sulphates), binwidth = 0.05) + 
  scale_x_continuous(breaks = seq(0, 2, 0.5)) +
  coord_cartesian(xlim = c(0, 2))

g12 <- g_base +
  geom_histogram(aes(x = alcohol), binwidth = 0.25) + 
  scale_x_continuous(breaks = seq(8, 15, 2)) +
  coord_cartesian(xlim = c(8, 15))

grid.arrange(g1, g2, g3, g4, g5, g6, 
             g7, g8, g9, g10, g11, g12, ncol=3)
```

There are some really interesting variations in the distributions here. Looking
closer at a few of the more interesting ones might prove quite valuable. 
Working from top-left to right, selected plots are analysed.

```{r echo=FALSE, warning=FALSE, message=FALSE, single_variable_hist}
base_hist <- ggplot(
  data = wine,
  aes(color=I('black'), fill=I('#990000'))
) 
```

### Acidity
```{r echo=FALSE, acidity_plot}
ac1 <- base_hist +
  geom_histogram(aes(x = fixed.acidity), binwidth = 0.25) + 
  scale_x_continuous(breaks = seq(4, 16, 2)) + 
  coord_cartesian(xlim = c(4, 16))

ac2 <- base_hist +
  geom_histogram(aes(x = volatile.acidity), binwidth = 0.05) + 
  scale_x_continuous(breaks = seq(0, 2, 0.5)) +
  coord_cartesian(xlim = c(0, 2))

grid.arrange(ac1, ac2, nrow=2)
```

**Fixed acidity** is determined by aids that do not evaporate easily -- 
tartaricacid. It contributes to many other attributes, including the taste, pH, 
color, and stability to oxidation, i.e., prevent the wine from tasting flat. 
On theother hand, **volatile acidity** is responsible for the sour taste in 
wine. A very high value can lead to sour tasting wine, a low value can make 
the wine seem heavy. 
(References: [1](http://waterhouse.ucdavis.edu/whats-in-wine/fixed-acidity), [2](http://waterhouse.ucdavis.edu/whats-in-wine/volatile-acidity).

```{r echo=FALSE, warning=FALSE, message=FALSE, acidity_univariate}
ac1 <- base_hist +
  geom_histogram(aes(x = fixed.acidity), binwidth = 0.25) + 
  scale_x_continuous(breaks = seq(4, 16, 2)) + 
  coord_cartesian(xlim = c(4, 16))

ac2 <- base_hist +
  geom_histogram(aes(x = volatile.acidity), binwidth = 0.05) + 
  scale_x_continuous(breaks = seq(0, 2, 0.5)) +
  coord_cartesian(xlim = c(0, 2))

ac3 <- base_hist +
  geom_histogram(aes(x = total.acidity), binwidth = 0.25) + 
  scale_x_continuous(breaks = seq(0, 18, 2)) +
  coord_cartesian(xlim = c(0, 18))

grid.arrange(ac1, ac2, ac3, nrow=3)

print("Summary statistics of Fixed Acidity")
summary(wine$fixed.acidity)
print("Summary statistics of Volatile Acidity")
summary(wine$volatile.acidity)
print("Summary statistics of Total Acidity")
summary(wine$total.acidity)
```

Of the wines we have in our dataset, we can see that most have a fixed acidity 
of 7.5. The median fixed acidity is 7.9, and the mean is 8.32. There is a 
slight skew in the data because a few wines possess a very high fixed acidity. 
The median volatile acidity is 0.52 g/dm^3, and the mean is 0.5278 g/dm^3. *It 
will be interesting to note which quality of wine is correlated to what level 
of acidity in the bivariate section.*

### Citric Acid 
Citric acid is part of the fixed acid content of most wines. A non-volatile 
acid, citric also adds much of the same characteristics as tartaric acid does. 
Again, here I would guess most good wines have a balanced amount of citric 
acid.

```{r echo=FALSE, warning=FALSE, message=FALSE, citric_acid_univariate}
base_hist +
  geom_histogram(aes(x = citric.acid), binwidth = 0.05) + 
  scale_x_continuous(breaks = seq(0, 1, 0.2)) + 
  coord_cartesian(xlim = c(0, 1))

print("Summary statistics of Citric Acid")
summary(wine$citric.acid)
print('Number of Zero Values')
table(wine$citric.acid == 0)
```

There is a very high count of zero in citric acid. To check if this is 
genuinely zero or merely a ‘not available’ value. A quick check using table 
function shows that there are 132 observations of zero values and no NA value 
in reported citric acid concentration. The citric acid concentration could be 
too low and insignificant hence was reported as zero.

As far as content wise the wines have a median citric acid level of 
0.26 g/dm^3, and a mean level of 0.271 g/dm^3.

### Sulfur-Dioxide & Sulphates
**Free sulfur dioxide** is the free form of SO2 exists in equilibrium between 
molecular SO2 (as a dissolved gas) and bisulfite ion; it prevents microbial 
growth and the oxidation of wine. **Sulphates** is a wine additive which can 
contribute to sulfur dioxide gas (SO2) levels, which acts as an anti-microbial 
moreover, antioxidant -- *overall keeping the wine, fresh*.

```{r echo=FALSE, warning=FALSE, message=FALSE, sulfur_univariate}
sul1 <- base_hist + geom_histogram(aes(x = free.sulfur.dioxide))
sul2 <- base_hist + geom_histogram(aes(x = log10(free.sulfur.dioxide)))

sul3 <- base_hist + geom_histogram(aes(x = total.sulfur.dioxide))
sul4 <- base_hist + geom_histogram(aes(x = log10(total.sulfur.dioxide)))

sul5 <- base_hist + geom_histogram(aes(x = sulphates))
sul6 <- base_hist + geom_histogram(aes(x = log10(sulphates)))

grid.arrange(sul1, sul2, sul3, sul4, sul5, sul6, nrow=3)
```

The distributions of all three values are positively skewed with a long tail. 
Thelog-transformation results in a normal-behaving distribution for 'total 
sulfur dioxide' and 'sulphates'. 

### Alcohol
Alcohol is what adds that special something that turns rotten grape juice 
into a drink many people love. Hence, by intuitive understanding, it should
be crucial in determining the wine quality.

```{r echo=FALSE, warning=FALSE, message=FALSE, alcohol_univariate}
base_hist +
  geom_histogram(aes(x = alcohol), binwidth = 0.25) + 
  scale_x_continuous(breaks = seq(8, 15, 2)) +
  coord_cartesian(xlim = c(8, 15))

print("Summary statistics for alcohol %age.")
summary(wine$alcohol)
```

The mean alcohol content for our wines is 10.42%, the median is 10.2%

### Quality
```{r echo=FALSE,warning=FALSE, message=FALSE, quality_univariate}
qplot(x=quality, data=wine, geom='bar',
      fill=I("#990000"), 
      col=I("black"))

print("Summary statistics - Wine Quality.")
summary(wine$quality)
```

Overall wine quality, rated on a scale from 1 to 10, has a normal shape and 
very few exceptionally high or low-quality ratings.

It can be seen that the minimum rating is 3 and 8 is the maximum for quality.
Hence, a variable called ‘rating’ is created based on variable quality.

* 8 to 7 are Rated A.

* 6 to 5 are Rated B.

* 3 to 4 are Rated C.

```{r echo=FALSE, quality_rating}
# Dividing the quality into 3 rating levels
wine$rating <- ifelse(wine$quality < 5, 'C', 
                      ifelse(wine$quality < 7, 'B', 'A'))

# Changing it into an ordered factor
wine$rating <- ordered(wine$rating,
                     levels = c('C', 'B', 'A'))

summary(wine$rating)

qr1 <- ggplot(aes(as.numeric(quality), fill=rating), data=wine) +
  geom_bar() +
  ggtitle ("Barchart of Quality with Rating") +
  scale_x_continuous(breaks=seq(3,8,1)) + 
  xlab("Quality") + 
  theme_pander() + scale_colour_few()

qr2 <- qplot(x=rating, data=wine, geom='bar',
      fill=I("#990000"), 
      col=I("black")) +
  xlab("Rating") +
  ggtitle("Barchart of Rating") + 
  theme_pander() 

grid.arrange(qr1, qr2, ncol=2)
```

The distribution of 'rating' is much higher on the 'B' rating wine as
seen in quality distribution. This is likely to cause overplotting. Therefore, 
a comparison of only the 'C' and 'A' wines is done to find distinctive 
properties that separate these two. The comparison is made using summary
statistics.

```{r echo=FALSE, rating_comparison}
print("Summary statistics of Wine with Rating 'A'")
summary(subset(wine, rating=='A'))

print("Summary statistics of Wine with Rating 'C'")
summary(subset(wine, rating=='C'))
```

On comparing the *mean statistic* of different attribute for 'A-rated' and 
'C-rated' wines (A → C), the following %age change is noted.

1. `fixed.acidity`: mean reduced by 11%.

2. `volatile.acidity` - mean increased by 80%.

3. `citric.acidity` - mean increased by 117%.

4. `sulphates` - mean reduced by 20.3%

5. `alcohol` - mean reduced by 12.7%.

6. `residualsugar` and `chloride` showed a very low variation.

These changes are, however, only suitable for estimation of important quality 
impacting variables and setting a way for further analysis. No conclusion
can be drawn from it.

## Univariate Analysis - Summary

### Overview
The red wine dataset features 1599 separate observations, each for a different 
red wine sample. As presented, each wine sample is provided as a single row in 
the dataset. Due to the nature of how some measurements are gathered, some 
values given represent *components* of a measurement total.  

For example, `data.fixed.acidity` and `data.volatile.acidity` are both obtained 
via separate measurement techniques, and must be summed to indicate the total 
acidity present in a wine sample. For these cases, I supplemented the data 
given by computing the total and storing in the data frame with a 
`data.total.*` variable. 

### Features of Interest
An interesting measurement here is the wine `quality`. It is the
subjective measurement of how attractive the wine might be to a consumer. The
goal here will is to try and correlate non-subjective wine properties with its 
quality.

I am curious about a few trends in particular --  **Sulphates vs. Quality** as 
low sulphate wine has a reputation for not causing hangovers, 
**Acidity vs. Quality** - Given that it impacts many factors like pH, 
taste, color, it is compelling to see if it affects the quality.
**Alcohol vs. Quality** - Just an interesting measurement. 

At first, the lack of an *age* metric was surprising since it is commonly
a factor in quick assumptions of wine quality. However, since the actual effect
of wine age is on the wine's measurable chemical properties, its exclusion here
might not be necessary.

### Distributions
 Many measurements that were clustered close to zero had a positive skew 
(you cannot have negative percentages or amounts). Others such as `pH` and 
`total.acidity` and `quality` had normal looking distributions. 

The distributions studied in this section were primarily used to identify the
trends in variables present in the dataset. This helps in setting up a track
for moving towards bivariate and multivariate analysis.

## Bivariate Plots Section

```{r echo=FALSE, message=FALSE, warning=FALSE, correlation_plots}
ggcorr(wine, 
       size = 2.2, hjust = 0.8,
       low = "#4682B4", mid = "white", high = "#E74C3C")
```

**Observations from the correlation matrix.**

* Total Acidity is highly correlatable with fixed acidity.

* pH appears correlatable with acidity, citric acid, chlorides, and residual
sugars.

* No single property appears to correlate with quality.

Further, in this section, metrics of interest are evaluated to check their
significance on the wine quality. Moreover, bivariate relationships between
other variables are also studied.

### Acidity vs. Rating & Quality

```{r echo=FALSE, message=FALSE, warning=FALSE, acidity_rating}

aq1 <- ggplot(aes(x=rating, y=total.acidity), data = wine) +
  geom_boxplot(fill = '#ffeeee') +
  coord_cartesian(ylim=c(0, quantile(wine$total.acidity, 0.99))) +
  geom_point(stat='summary', fun.y=mean,color='red') + 
  xlab('Rating') + ylab('Total Acidity')
  
aq2 <- ggplot(aes(x=quality, y=total.acidity), data = wine) +
  geom_boxplot(fill = '#ffeeee') +
  coord_cartesian(ylim=c(0, quantile(wine$total.acidity, 0.99))) +
  geom_point(stat='summary', fun.y=mean, color='red') +
  xlab('Quality') + ylab('Total Acidity') + 
  geom_jitter(alpha=1/10, color='#990000') + 
  ggtitle("\n")
  
grid.arrange(aq1, aq2, ncol=1)
```

The boxplots depicting quality also depicts the distribution 
of various wines, and we can again see 5 and 6 quality wines have the most 
share. The blue dot is the mean, and the middle line shows the median.

The box plots show how the acidity decreases as the quality of wine improve. 
However, the difference is not very noticeable. Since most wines tend to 
maintain a similar acidity level & given the fact that *volatile acidity* is
responsible for the sour taste in wine, hence a density plot of the said 
attribute is plotted to investigate the data.

```{r echo=FALSE, message=FALSE, warning=FALSE, acidity_quality_rating}
ggplot(aes(x = volatile.acidity, fill = quality, color = quality), 
       data = wine) + 
  geom_density(alpha=0.08)
```

Red Wine of `quality` 7 and 8 have their peaks for `volatile.acidity` well 
below the 0.4 mark. Wine with `quality` 3 has the pick at the most right 
hand side (towards more volatile acidity). This shows that the better quality
wines are lesser sour and in general have lesser acidity. 

### Alcohol vs. Quality

```{r echo=FALSE, message=FALSE, warning=FALSE, alcohol_quality_sugar}
qas0 <- ggplot(aes(x=alcohol, y=as.numeric(quality)), data=wine) + 
  geom_jitter(alpha=1/12) + 
  geom_smooth() + 
  ggtitle("Alcohol Content vs. Quality") + 
  ylab("Quality") + xlab("Alcohol")

qas1 <- ggplot(aes(x=alcohol), data=wine) + 
  geom_density(fill=I("#BB0000")) + 
  facet_wrap("quality") + 
  ggtitle("Alcohol Content for \nWine Quality Ratings") +
  ylab("Density") + xlab("Alcohol")

qas2 <- ggplot(aes(x=residual.sugar, y=alcohol), data=wine) +
  geom_jitter(alpha=1/12) + 
  geom_smooth() +
  ggtitle("Alcohol vs. Residual Sugar Content") + 
  ylab("Alcohol") + xlab("Residual Sugar")

grid.arrange(qas1, arrangeGrob(qas0, qas2), ncol=2)
```

The plot between residual sugar and alcohol content suggests that there is no 
erratic relation between sugar and alcohol content, which is surprising as 
alcohol is a byproduct of the yeast feeding off of sugar during the 
fermentation process. That inference could not be established here.

Alcohol and quality appear to be somewhat correlatable. Lower quality wines 
tend to have lower alcohol content. This can be further studied using boxplots.

```{r echo=FALSE, message=FALSE, warning=FALSE}

quality_groups <- group_by(wine, alcohol)

wine.quality_groups <- summarize(quality_groups,
                          acidity_mean = mean(volatile.acidity),
                          pH_mean = mean(pH),
                          sulphates_mean = mean(sulphates),
                          qmean = mean(as.numeric(quality)),
                          n = n())

wine.quality_groups <- arrange(wine.quality_groups, alcohol)
```

```{r echo=FALSE, message=FALSE, warning=FALSE, alcohol_quality}
ggplot(aes(y=alcohol, x=factor(quality)), data = wine) +
  geom_boxplot(fill = '#ffeeee')+
  xlab('quality')
```

The boxplots show an indication that higher quality wines have higher alcohol 
content. This trend is shown by all the quality grades from 3 to 8 except 
quality grade 5.

**Does this mean that by adding more alcohol, we'd get better wine?**

```{r echo=FALSE, message=FALSE, warning=FALSE}
ggplot(aes(alcohol, qmean), data=wine.quality_groups) +
  geom_smooth() +
  ylab("Quality Mean") +
  scale_x_continuous(breaks = seq(0, 15, 0.5)) +
  xlab("Alcohol %")
```

The above line plot indicates nearly a linear increase till 13% alcohol 
concetration, followed by a steep downwards trend. The graph has to be 
smoothened to remove variances and noise.

### Sulphates vs. Quality

```{r echo=FALSE, message=FALSE, warning=FALSE, sulphates_quality}
ggplot(aes(y=sulphates, x=quality), data=wine) +
  geom_boxplot(fill="#ffeeee")
```

Good wines have higher sulphates values than bad wines, though the difference 
is not that wide.

```{r echo=FALSE, message=FALSE, warning=FALSE, sulphates_qplots}
sq1 <- ggplot(aes(x=sulphates, y=as.numeric(quality)), data=wine) +
  geom_jitter(alpha=1/10) +
  geom_smooth() +
  xlab("Sulphates") + ylab("Quality") + 
  ggtitle("Sulphates vs. Quality")

sq2 <- ggplot(aes(x=sulphates, y=as.numeric(quality)), 
              data=subset(wine, wine$sulphates < 1)) +
  geom_jitter(alpha=1/10) +
  geom_smooth() +
  xlab("Sulphates") + ylab("Quality") + 
  ggtitle("\nSulphates vs Quality without Outliers") 

grid.arrange(sq1, sq2, nrow = 2)
```

There is a slight trend implying a relationship between sulphates and wine
quality, mainly if extreme sulphate values are ignored, i.e., because 
disregarding measurements where sulphates > 1.0 is the same as disregarding
the positive tail of the distribution, keeping just the normal-looking portion.
However, the relationship is mathematically, still weak.

## Bivariate Analysis - Summary

There is no apparent and mathematically strong correlation between any wine 
property and the given quality. Alcohol content is a strong contender, but even 
so, the correlation was not particularly strong.

Most properties have roughly normal distributions, with some skew in one tail. 
Scatterplot relationships between these properties often showed a slight trend 
within the bulk of property values. However, as soon as we leave the 
expected range, the trends reverse. For example, Alcohol Content or 
Sulphate vs. Quality. The trend is not a definitive one, but it is seen in 
different variables. 

Possibly, obtaining an outlier property (say sulphate content) is particularly 
challenging to do in the wine making process. Alternatively, there is a change 
that the wines that exhibit outlier properties are deliberately of a 
non-standard variety. In that case, it could be that wine judges have a harder 
time agreeing on a quality rating.

## Multivariate Plots Section

This section includes visualizations that take bivariate analysis a step 
further, i.e., understand the earlier patterns better or to strengthen the 
arguments that were presented in the previous section.

### Alcohol, Volatile Acid & Wine Rating

```{r echo=FALSE, message=FALSE, warning=FALSE, alcohol_acid_quality}
ggplot(wine, aes(x=alcohol, y=volatile.acidity, color=quality)) +
  geom_jitter(alpha=0.8, position = position_jitter()) +
  geom_smooth(method="lm", se = FALSE, size=1) + 
  scale_color_brewer(type='seq',
                   guide=guide_legend(title='Quality')) +
  theme_pander()
```

Earlier inspections suggested that the volatile acidity and alcohol had high 
correlations values of negative and positive. Alcohol seems to vary more than 
volatile acidity when we talk about quality, nearly every Rating A wine has 
less than 0.6 volatile acidity.

### Understanding the Significance of Acidity

```{r echo=FALSE, message=FALSE, warning=FALSE, acid_quality}
ggplot(subset(wine, rating=='A'|rating=='C'),
       aes(x=volatile.acidity, y=citric.acid)) +
  geom_point() + 
  geom_jitter(position=position_jitter(), aes(color=rating)) +
  geom_vline(xintercept=c(0.6), linetype='dashed', size=1, color='black') +
  geom_hline(yintercept=c(0.5), linetype='dashed', size=1, color='black') +
  scale_x_continuous(breaks = seq(0, 1.6, .1)) +
  theme_pander() + scale_colour_few()
```

Nearly every wine has volatile acidity less than 0.8. As discussed earlier the
A rating wines all have volatile.acidity of less than 0.6. For wines with
rating B, the volatile acidity is between 0.4 and 0.8. Some C rating wine have
a volatile acidity value of more than 0.8

Most A rating wines have citric acid value of 0.25 to 0.75 while the B rating 
wines have citric acid value below 0.50.

### Understanding the Significance of Sulphates

```{r echo=FALSE, message=FALSE, warning=FALSE}
ggplot(subset(wine, rating=='A'|rating=='C'), aes(x = alcohol, y = sulphates)) +
    geom_jitter(position = position_jitter(), aes(color=rating)) +
  geom_hline(yintercept=c(0.65), linetype='dashed', size=1, color='black') +
  theme_pander() + scale_colour_few() +
  scale_y_continuous(breaks = seq(0, 2, .2))
```

It is incredible to see that nearly all wines lie below 1.0 sulphates level. 
Due to overplotting, wines with rating B have been removed. It can be seen 
rating A wines mostly have sulphate values between 0.5 and 1 and the best rated 
wines have sulphate values between 0.6 and 1. Alcohol has the same values as 
seen before.

### Density & Sugar

```{r echo=FALSE, message=FALSE, warning=FALSE, Multivariate_Plots2}
da1 <- ggplot(aes(x=density, y=total.acidity, color=as.numeric(quality)), 
              data=wine) + 
  geom_point(position='jitter') +
  geom_smooth() +
  labs(x="Total Acidity", y="Density", color="Quality") +
  ggtitle("Density vs. Acidity Colored by Wine Quality Ratings")

cs2 <- ggplot(aes(x=residual.sugar, y=density, color=as.numeric(quality)), 
              data=wine) + 
  geom_point(position='jitter') +
  geom_smooth() + 
  labs(x="Residual Sugar", y="Density", color="Quality") +
  ggtitle("\nSugar vs. Chlorides colored by Wine Quality Ratings")

grid.arrange(da1, cs2)
```

Higher quality wines appear to have a slight correlation with higher acidity
across all densities. Moreover, there are abnormally high and low quality wines 
coincident with higher-than-usual sugar content. 

## Multivariate Analysis - Summary
Based on the investigation, it can be said that higher `citric.acid` and 
lower `volatile.acidity` contribute towards better wines. Also, better wines 
tend to have higher alcohol content. 

There were surprising results with `suplhates` and `alcohol` graphs. 
Sulphates had a better correlation with quality than citric acid, still the 
distribution was not that distinct between the different quality wines. Further 
nearly all wines had a sulphate content of less than 1, irrespective of the 
alcohol content; suplhate is a byproduct of fermantation just like 
alcohol. 

Based on the analysis presented, it can be noted because wine rating is a 
subjective measure, it is why statistical correlation values are not a very 
suitable metric to find important factors. This was realized half-way through
the study. The graphs aptly depict that there is a suitable range and it is 
some combination of chemical factors that contribute to the flavour of wine.

## Final Plots and Summary

### Plot One

```{r echo=FALSE, message=FALSE, warning=FALSE, plot_2}
qr1 <- ggplot(aes(as.numeric(quality), fill=rating), data=wine) +
  geom_bar() +
  ggtitle ("Barchart of Quality with Rating") +
  scale_x_continuous(breaks=seq(3,8,1)) + 
  xlab("Quality") + 
  theme_pander() + scale_colour_few()

qr2 <- qplot(x=rating, data=wine, geom='bar',
      fill=I("#990000"), 
      col=I("black")) +
  xlab("Rating") +
  ggtitle("Barchart of Rating") + 
  theme_pander() 

grid.arrange(qr1, qr2, ncol=2)
```

#### Description One
The plot is from the univariate section, which introduced the idea of 
this analysis. As in the analysis, there are plenty of visualizations which 
only plot data-points from A and C rated wines. A first comparison of only 
the 'C' and 'A' wines helped find distinctive properties that separate these 
two.

It also suggests that it is likely that the critics can be highly subjective as
they do not rate any wine with a measure of 1, 2 or 9, 10. With most wines 
being mediocre, the wines that had the less popular rating must've caught the 
attention of the wine experts, hence, the idea was derived to compare these two
rating classes.

### Plot Two

```{r, echo=FALSE, warning=FALSE, message=FALSE, plot_1a}
ggplot(aes(x=alcohol), data=wine) + 
  geom_density(fill=I("#BB0000")) + 
  facet_wrap("quality") + 
  ggtitle("Alcohol Content for Wine Quality Ratings") +
  labs(x="Alcohol [%age]", y="") +
  theme(plot.title = element_text(face="plain"),
        axis.title.x = element_text(size=10),
        axis.title.y = element_text(size=10))
```

```{r echo=FALSE, message=FALSE, warning=FALSE, plot_1b}
fp1 <- ggplot(aes(y=alcohol, x=quality), data = wine)+
  geom_boxplot() +
  xlab('Quality') +
  ylab("Alcohol in % by Volume") +
  labs(x="Quality", y="Alcohol [%age]") +
  ggtitle("Boxplot of Alcohol and Quality") +
  theme(plot.title = element_text(face="plain"),
        axis.title.x = element_text(size=10),
        axis.title.y = element_text(size=10))

fp2 <-ggplot(aes(alcohol, qmean), data=wine.quality_groups) +
  geom_smooth() +
  scale_x_continuous(breaks = seq(0, 15, 0.5)) +
  ggtitle("\nLine Plot of Quality Mean & Alcohol Percentage") + 
  labs(x="Alcohol [%age]", y="Quality (Mean)") +
  theme(plot.title = element_text(face="plain"),
        axis.title.x = element_text(size=10),
        axis.title.y = element_text(size=10))

grid.arrange(fp1, fp2)
```

#### Description Two

These are plots taken from bivariate analysis section discussing the effect of 
alcohol percentage on quality.

The first visualization was especially appealing to me because of the way that 
you can almost see the distribution shift from left to right as wine ratings 
increase. Again, just showing a general tendency instead of a substantial
significance in judging wine quality.

The above boxplots show a steady rise in the level of alcohol. An interesting 
trend of a decrement of quality above 13%, alcohol gave way to further analysis 
which shows that a general correlation measure might not be suitable for the 
study.

The plot that follows set the basis for which I carried out the complete 
analysis. Rather than emphasizing on mathematical correlation measures, the 
inferences drawn were based on investigating the visualizations. This felt 
suitable due to the subjectivity in the measure of wine quality.

### Plot Three

```{r echo=FALSE, messages=FALSE, warning=FALSE, plot_3}
fp3 <- ggplot(subset(wine, rating=='A'|rating=='C'), 
              aes(x = volatile.acidity, y = citric.acid)) +
  geom_point() + 
  geom_jitter(position=position_jitter(), aes(color=rating)) +
  geom_vline(xintercept=c(0.6), linetype='dashed', size=1, color='black') +
  geom_hline(yintercept=c(0.5), linetype='dashed', size=1, color='black') +
  scale_x_continuous(breaks = seq(0, 1.6, .1)) +
  theme_pander() + scale_colour_few() +
  ggtitle("Wine Rating vs. Acids") + 
  labs(x="Volatile Acidity (g/dm^3)", y="Citric Acid (g/dm^3)") +
  theme(plot.title = element_text(face="plain"),
        axis.title.x = element_text(size=10),
        axis.title.y = element_text(size=10),
        legend.title = element_text(size=10))

fp4 <- ggplot(subset(wine, rating=='A'|rating=='C'), 
              aes(x = alcohol, y = sulphates)) +
  geom_jitter(position = position_jitter(), aes(color=rating)) +
  geom_hline(yintercept=c(0.65), linetype='dashed', size=1, color='black') +
  theme_pander() + scale_colour_few() +
  scale_y_continuous(breaks = seq(0,2,.2)) +
  ggtitle("\nSulphates, Alcohol & Wine-Rating") + 
  labs(x="Alcohol [%]", y="Sulphates (g/dm^3)") +
  theme(plot.title = element_text(face="plain"),
        axis.title.x = element_text(size=10),
        axis.title.y = element_text(size=10),
        legend.title = element_text(size=10))

grid.arrange(fp3, fp4, nrow=2)
```

#### Description Three
These plots served as finding distinguishing boundaries for given attributes,
i.e., `sulphates`, `citric.acid`, `alcohol`, `volatile.acidity`. The 
conclusions drawn from these plots are that sulphates should be high but less 
than 1 with an alcohol concentration around 12-13%, along with less (< 0.6) 
volatile acidity. It can be viewed nearlyas a depiction of a classification 
methodology without application of any machine learning algorithm. Moreover, 
these plots strengthened the arguments laid in the earlier analysis of the data.

------

## Reflection
In this project, I was able to examine relationship between *physicochemical* 
properties and identify the key variables that determine red wine quality, 
which are alcohol content volatile acidity and sulphate levels.

The dataset is quite interesting, though limited in large-scale implications. 
I believe if this dataset held only one additional variable it would be vastly 
more useful to the layman. If *price* were supplied along with this data 
one could target the best wines within price categories, and what aspects 
correlated to a high performing wine in any price bracket. 

Overall, I was initially surprised by the seemingly dispersed nature of the 
wine data. Nothing was immediately correlatable to being an inherent quality
of good wines. However, upon reflection, this is a sensible finding. Wine
making is still something of a science and an art, and if there was one
single property or process that continually yielded high quality wines, the
field wouldn't be what it is. 

According to the study, it can be concluded that the best kind of wines are the
ones with an alcohol concentration of about 13%, with low volatile acidity & 
high sulphates level (with an upper cap of 1.0 g/dm^3).

### Future Work & Limitations
With my amateurish knowledge of wine-tasting, I tried my best to relate it to
how I would rate a bottle of wine at dining. However, in the future, I would 
like to do some research into the winemaking process. Some winemakers might 
actively try for some property values or combinations, and be finding those 
combinations (of 3 or more properties) might be the key to truly predicting 
wine quality. This investigation was not able to find a robust generalized 
model that would consistently be able to predict wine quality with any degree 
of certainty.

If I were to continue further into this specific dataset, I would aim to 
train a classifier to correctly predict the wine category, in order to better 
grasp the minuteness of what makes a good wine. 

Additionally, having the wine type would be helpful for further analysis. 
Sommeliers might prefer certain types of wines to have different 
properties and behaviors. For example, a Port (as sweet desert wine) 
surely is rated differently from a dark and robust abernet Sauvignon, 
which is rated differently from a bright and fruity Syrah. Without knowing
the type of wine, it is entirely possible that we are almost literally
comparing apples to oranges and can't find a correlation.