wineQualityWhite-EDA.rmd

Exploratory Data Analysis of White Wine Quality by HINDHUJA GUTHA
==================================================================

```{r echo=FALSE, message=FALSE, warning=FALSE, packages_and_global_options}
# Load packages used in this analysis
library(ggplot2)
library(gridExtra)
library(ggthemes)
knitr::opts_chunk$set(fig.width=8,fig.height=4,fig.path='Figs/',
                      fig.align='center',tidy=TRUE,
                      echo=FALSE,warning=FALSE,message=FALSE)
```

```{r echo=FALSE, Load_the_Data}
# Load the Data
whiteWineData <- read.csv('wineQualityWhites.csv')
```

# Introduction:
 The dataset used in this EDA is related to white wine samples of the Portuguese "Vinho Verde" wine.For more details, consult: http://www.vinhoverde.pt/en/ or the reference [Cortez et al., 2009].

 Due to privacy and logistic issues, only physicochemical (inputs) and sensory (the output) variables are available (e.g. there is no data about grape types, wine brand, wine selling price, etc.).

 The inputs include objective tests (e.g. PH values) and the output is based on sensory data (median of at least 3 evaluations made by wine experts). Each expert graded the wine quality between 0 (very bad) and 10 (very excellent).

 The classes are ordered and not balanced (e.g. there are munch more normal wines than excellent or poor ones). Outlier detection algorithms could be used to detect the few excellent or poor wines.

**Attribute Information :**

**Input variables (based on physicochemical tests):**

* 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)

# Univariate Plots Section

In this section summary of all variables and information about dataset is 
analysed along with histograms for important variables and 
if necessary new variables are created

##White Wine Dataset Summary

**Null values in Dataset**

```{r echo=FALSE, message=FALSE, warning=FALSE,check_null}
#Check for null values in dataset
sum(is.na(whiteWineData))
```

**row count**

```{r echo=FALSE, message=FALSE, warning=FALSE,check_row_count}
#Check row count
nrow(whiteWineData)
```

**column count**

```{r echo=FALSE, message=FALSE, warning=FALSE,check_column_count}
#Check column count
ncol(whiteWineData)
```

**Dataset Summary**

```{r echo=FALSE, message=FALSE, warning=FALSE,dataset_sumary}
#Get summary of dataset
summary(whiteWineData)
```

**Dataset Observations:**

* Datset has 13 columns . X refers to the number of the sampled white wine.
* Other columns are attributes for sample X
* volatile.acidity (0.08-1.10) range is less than fixed.acidity (3.8 to 14.2) 
* Some samples doesn't contain citric acid which suggests it's used for flavor.
* Maximum value of sugar(65.8) is much greater than mean and median
* free sulphur dioxide also has max value 289 compared to mean value 35.31
* Density of all samples are almost consistent 
* pH range is in between 2.720-3.820
* Minimum alcohol is 8% whereas Maximum is 14.20%
* In general avg alcohol is around 11-13% but it can also vary from 5.5-20% 
which suggests that samples in dataset has alcohol in required range.
* None of the samples are rated very bad(score 0) or very excellent(score 10)
* Maximum rated wine quality is 9. None of the sample is rated 10.

**idealpH categorical variable**
 
 Based on the information 3-3.4 is best pH level for white wines, a categorical
variable idealPh is created which takes value 'Yes' when pH level is in between 3-3.4 
otherwise the value will be 'No'

```{r echo=FALSE, message=FALSE, warning=FALSE,create_ph_categorical_variable}
#Create categorical variable idealpH with 2 Factor levels Yes/No
#'Yes' for pH range from 3-3.4 otherwise 'No'
whiteWineData$idealpH  <- factor(ifelse(whiteWineData$pH >= 3 & 
                                      whiteWineData$pH <= 3.4 , "Yes", "No"))

```


```{r echo=FALSE, message=FALSE, warning=FALSE,create_hist_function}
# Create Function to print Univariate Plots with input args-> attribute name , 
#x-axis lower limit ,x-axis Upper limit,binwidth(default 0.5),break(default 1) 
univariatePlots <- function(curAttr,lowerL,upperL,bw = 0.5, brk = 1){
qplot(whiteWineData[, c(curAttr)],  col=I("lightseagreen"),
          fill=I("lightblue"), binwidth = bw) +
scale_x_continuous(breaks = seq(lowerL,upperL,brk), lim = c(lowerL,upperL))+
      labs(title=paste("Histogram for", curAttr), x=curAttr, y="Count")
}
```

**Fixed Acidity Summary**

```{r echo=FALSE,message=FALSE, warning=FALSE,fixed_acidity_summary}
#Get Summary of variable fixed.acidity
summary(whiteWineData$fixed.acidity)
```

**Fixed Acidity plot**

```{r echo=FALSE,message=FALSE, warning=FALSE,fixed_acidity_plot}
#Histogram for fixed.acidity using function univariatePlots
univariatePlots("fixed.acidity",3,15)
```

* fixed.acidity is normally distributed
* As fixed.acidity value increased from 6.75 there is a decrease in sample count

**Volatile Acidity Summary**

```{r echo=FALSE,message=FALSE, warning=FALSE,volatile_acidity_summary}
#Get Summary of variable volatile.acidity
summary(whiteWineData$volatile.acidity)
```

**Volatile Acidity  plot**

```{r echo=FALSE,message=FALSE, warning=FALSE,volatile_acidity_plot}
#Histogram for volatile.acidity using function univariatePlots
univariatePlots("volatile.acidity",0.08,1.10,0.04,0.1)
```

* Volatile acidity has some outliers and ditribution is slightly right skewed
* majority samples have volatile acidity below 0.4

**Citric Acid Summary**

```{r echo=FALSE,message=FALSE, warning=FALSE,citric_acid_summary}
#Get Summary of variable citric.acid
summary(whiteWineData$citric.acid)
```

**Citric Acid  plot**

```{r echo=FALSE,message=FALSE, warning=FALSE,citric_acid_plot}
#Histogram for citric.acid using function univariatePlots
univariatePlots("citric.acid",0.00,1.68,0.05,0.4)
```

* citric.acid has some major outliers and distribution is normal

**Residual Sugar Summary**

```{r echo=FALSE,message=FALSE, warning=FALSE,residual_sugar_summary}
#Get Summary of variable residual.sugar
summary(whiteWineData$residual.sugar)
```

**Residual Sugar plot**

```{r echo=FALSE,message=FALSE, warning=FALSE,residual_sugar_plot}
#Histogram for residual.sugar using function univariatePlots
univariatePlots("residual.sugar",0,66,3,5)
```

* Distribution of residual sugar is right skewed
* There seems to be a negative association between count and residual sugar
* As residual.sugar value increased , there is decrease in sample count, which
means majority samples have less residual sugar.

**Total sulfur dioxide Summary**

```{r echo=FALSE,message=FALSE, warning=FALSE,total_sulfur_dioxide_summary}
#Get Summary of variable total.sulfur.dioxide
summary(whiteWineData$total.sulfur.dioxide)
```

**Total Sulfur dioxide  plot**

```{r echo=FALSE,message=FALSE, warning=FALSE,total_sulfur_dioxide_plot}
#Histogram for total.sulfur.dioxide using function univariatePlots
univariatePlots("total.sulfur.dioxide",0,450,15,30)
```

* Total sulfur dioxide is normally distributed but there are some outliers
* Majority samples have total sulfur dioxide below 200

**Density Summary**

```{r echo=FALSE,message=FALSE, warning=FALSE,density_summary}
#Get Summary of variable density
summary(whiteWineData$density)
```

**Density plot**

```{r echo=FALSE,message=FALSE, warning=FALSE,density_plot}
#Histogram for density using function univariatePlots
univariatePlots("density",0.98,1.04,0.002,0.01)
```

* density is normally distributed with some major outliers

**pH Summary**

```{r echo=FALSE,message=FALSE, warning=FALSE,pH_summary}
#Get Summary of variable pH
summary(whiteWineData$pH)
```

**pH plot**

```{r echo=FALSE,message=FALSE, warning=FALSE,pH_plot}
#Histogram for pH using function univariatePlots
univariatePlots("pH",2.7,3.9,0.04,0.1)
```

* pH range is normally distributed 
* 3 to 3.3 seems to be popular pH range

**IdealpH category variable (3-3.4pH value) Summary**

```{r echo=FALSE,message=FALSE, warning=FALSE,idealPH_summary}
#Get Summary of variable idealpH
summary(whiteWineData$idealpH)
```

**Bar plot for idealPH variable**

```{r echo=FALSE,message=FALSE, warning=FALSE,idealPH_plot}
#Create bar plot for idealpH variable
ggplot(data=whiteWineData, 
  aes(x=idealpH, col=I("lightseagreen"), fill=I("lightblue"))) + 
geom_bar(aes(y = (..count..)*100/sum(..count..)))+
labs(title=paste("Bar plot for Ideal pH"), x="idealpH", y="Total Percentage(%)")

```

* More than 80% samples are in idealPH range

**Sulphates Summary**

```{r echo=FALSE,message=FALSE, warning=FALSE,sulphates_summary}
#Get Summary of variable sulphates
summary(whiteWineData$sulphates)
```

**Sulphates  plot**

```{r echo=FALSE,message=FALSE, warning=FALSE,sulphates_plot}
#Histogram for sulphates using function univariatePlots
univariatePlots("sulphates",0.2,1.2,0.05,0.1)
```

* Sulphates distribution seems to be right skewed

**Alcohol(% by volume) Summary**

```{r echo=FALSE,message=FALSE, warning=FALSE,alcohol_summary}
#Get Summary of variable alcohol
summary(whiteWineData$alcohol)
```

**Alcohol(% by volume) plot**

```{r echo=FALSE,message=FALSE, warning=FALSE,alcohol_plot}
#Histogram for alcohol using function univariatePlots
univariatePlots("alcohol",8,15,0.2)
```

* Distribution for Alcohol is little bit right skewed

**Quality Summary**

```{r echo=FALSE,message=FALSE, warning=FALSE,quality_summary}
#Get Summary of variable quality
summary(whiteWineData$quality)
```

**Quality plot**

```{r echo=FALSE,message=FALSE, warning=FALSE,quality_plot}
#Histogram for quality using function univariatePlots
univariatePlots("quality",0,10,1,1)
```

* Distribution of quality levels is left skewed 
* Majority samples have quality levels 5,6 & 7

# Univariate Analysis

**Number of Instances in white wine Dataset :** 4898. 

**Number of Attributes:** Total 13 columns,column "X" to represent sample &
 remaining 12 columns represent sample attributes

**Missing Attribute Values:** None

### What is the structure of your dataset?
dataset is tidy and there are no missing values . 

### What is/are the main feature(s) of interest in your dataset?
residual sugar, alcohol,pH and fixed.acidity are main attributes

### What other features in the dataset do you think will help support your \
investigation into your feature(s) of interest?
quality,sulphates and density can help in understanding more about wine

### Did you create any new variables from existing variables in the dataset?
I have created idealpH category variable based on ideal pH range 3-3.4

### Of the features you investigated, were there any unusual distributions? \
Did you perform any operations on the data to tidy, adjust, or change the form \
of the data? If so, why did you do this?
* Most of the individual variables are normally distributed.
* Residual sugar distribution is skewed.
* fixed.acidity,volatile.acidity,citric.acid,total.sulfur.dioxide,density and 
residual.sugar has some outliers
* More than 80% of samples are in ideal pH range
* All levels of quality are not present
 
# Bivariate Plots Section

 Based on above individual variable analysis ,in this section 
Bivariate Analysis is done to show comparisons and trends between two varaibles
 Scatterplot is a good way to analyze bivariate relationshhip , It is used.
Plots are analysed for below pairs

* fixed.acidity vs pH
* sulphates vs pH
* fixed.acidity vs sulphates
* total.sulphur.dioxide vs quality
* residual.sugar vs alcohol
* residual.sugar vs quality
* alcohol vs quality
* idealPh vs quality

```{r echo=FALSE,message=FALSE, warning=FALSE,fixed.acidity_and_pH}
#scatter plot for Fixed acidity vs pH 
ggplot(whiteWineData,aes(x = pH, y = fixed.acidity))+
geom_point() +
geom_jitter()+geom_smooth(stat = 'smooth', color = 'seagreen', method = 'gam',
formula = y ~ s(x, bs = "cs"))+labs(x = "pH value", y = "Fixed Acidity Value",
  title = "Fixed acidity vs pH")+
  theme_economist() + scale_color_economist()
```

* negative association between fixed acidity and pH value though it is not a
non-linear correlation.
* Increase in fixed acidity decreased pH value which suggests that fixed.acidity
contributes majorly to the overall pH of the sample

```{r echo=FALSE,message=FALSE, warning=FALSE,sulphates_and_pH}
#scatter plot for Sulphates vs pH
ggplot(whiteWineData,aes(x = pH, y = sulphates))+
geom_point() +
geom_jitter()+geom_smooth(stat = 'smooth', color = 'seagreen', method = 'gam',
formula = y ~ s(x, bs = "cs"))+labs(x = "pH value", y = "Sulphates value",
  title = "Sulphates vs pH")+
  theme_economist() + scale_color_economist()
```

* postivie association between sulphates value and pH value though it is not a
linear correlation.
* Increase in sulphates value increased pH value which mean sulphates is one of 
the major attributes in deciding the overall pH of the sample

```{r echo=FALSE,message=FALSE, warning=FALSE,sulphates_and_fixed_acidity}
#scatter plot for Fixed acidity vs Sulphates
ggplot(whiteWineData,aes(x = fixed.acidity, y = sulphates))+
geom_point() +
geom_jitter()+geom_smooth(stat = 'smooth', color = 'seagreen', method = 'gam',
formula = y ~ s(x, bs = "cs"))+labs(x = "Fixed Aciidty value",
y = "Sulphates value",title = "Fixed Acidity vs Sulphates")+
theme_economist() + scale_color_economist()
```

* From above plot,No association can be seen between fixed.acidity and sulphates
,Further analysis needs to be done to decide association
* Very few samples have fixed acidity above 9 
* similarly there are few samples with sulphates value above 0.8

```{r echo=FALSE,message=FALSE, warning=FALSE,so2_and_quality}
#scatter plot for Total sulfur dioxide vs Quality
ggplot(whiteWineData,aes(x = as.factor(quality), y = total.sulfur.dioxide))+
geom_point() +
geom_jitter()+labs(x = "Quality",
y = "Total Sulfur dioxide",title = "Total Sulfur dioxide vs Quality")+
theme_economist() 
```

* Majority wine samples have total sulfur dioxide above 11.
* Quality levels 5,6 & 7 seems to have more samples


```{r echo=FALSE,message=FALSE, warning=FALSE,sugar_and_alcohol}
#scatter plot for Residual Sugar vs Alcohol
ggplot(whiteWineData,aes(x = alcohol, y = residual.sugar))+
geom_point() +
geom_jitter()+geom_smooth(stat = 'smooth', color = 'green', method = 'gam',
formula = y ~ s(x, bs = "cs"))+labs(x = "Alcohol (% by Volume)",
y = "Residual sugar(grams)",title = "Residual Sugar vs Alcohol")+
  scale_color_brewer( name="Alcohol")+
theme_bw()
```

* There is negative association between residual sugar and alcohol %
* As sugar level decreased there is an increase in alcohol % and also decrease 
in sample count

```{r echo=FALSE,message=FALSE, warning=FALSE,residual_sugar_and_quality}
#scatter plot for Residual Sugar vs Quality 
ggplot(whiteWineData,aes(x = as.factor(quality), y = residual.sugar,
                         color=as.factor(quality)))+
geom_point() +geom_jitter()+labs(x = "Quality(0-10)",
y = "Residual sugar(grams)",title = "Residual Sugar vs Quality")+
    scale_color_brewer( name="Quality levels")+
theme_bw()
```

* There seems to be a decrease in sample count when residual.sugar is increased
* Majority samples have residual.sugar value below 20
* from quality level 5 , as quality level increased there is decrease in overall
sample count
* There is no direct association between residual sugar and quality

```{r echo=FALSE,message=FALSE, warning=FALSE,alcohol_and_quality}
#scatter plot for Alcohol vs Quality 
ggplot(whiteWineData,aes(x = as.factor(quality), y = alcohol,
                         color=as.factor(quality)))+
geom_point() +geom_jitter()+labs(x = "Quality(0-10)",
  y = "Alcohol(% by Volume)",title = "Alcohol vs Quality")+
  scale_color_brewer( name="Quality levels")+
theme_bw()
```

* Quality levels 5,6, & 7 have more samples
* Majority samples have alcohol level above 11%
 
```{r echo=FALSE,message=FALSE, warning=FALSE,idealph_quality}
#scatter plot for ideal pH vs Quality
ggplot(whiteWineData,aes(x = idealpH, y = as.factor(quality),
                color=idealpH))+
geom_point() +
geom_jitter()+labs(x = "Ideal pH value for white wine",y = "Quality(0-10)",
    title = "Ideal pH vs Quality")

```

* Quality depends on ideal pH range 3-3.4, as samples in ideal pH range have 
more count compared to samples which are not in ideal Ph range.
* This can be seen at all quality levels in above plot

# Bivariate Analysis

### Talk about some of the relationships you observed in this part of the \
investigation. How did the feature(s) of interest vary with other features in \
the dataset?
* Negative association between fixed.acidity and pH value
* Positive association between sulphate and pH value
* Trend increased and decreased multiple times for fixed.acidity and sulphates
* Most of the sample shave residual sugar below 20 grams with some outliers
* Alochol and quality seems to have a postive correlation 
* Majority samples have alcohol level above 10 % and are in ideal pH range

### Did you observe any interesting relationships between the other features \
(not the main feature(s) of interest)?

* Total sulfur dioxide is used to determine freshness of wine and
majority samples have total sulfur dioxide above 100 which suggests that 
most wine samples are not aged well.

### What was the strongest relationship you found?

* strongest relationship is found between fixed.acidity vs pH & sulphate vs pH

# Multivariate Plots Section

 In this section association between multiple variables is explored.
Based on Bivariate plots Analysis below variables are analyzed together.

* fixed.acidity,sulphates and pH value
* residual.sugar,alcohol and quality

 **Relationship between Fixed Acidity , Sulphates and pH value**

```{r echo=FALSE,message=FALSE, warning=FALSE,acidity_sulphates_and_pH}
#Scatter plot for Fixed Acidity , Sulphates and pH value
ggplot(whiteWineData,aes(x = sulphates, y = fixed.acidity ,
                         color= pH))+
geom_point(alpha=0.5, size = 1, shape = 1)+ 
  geom_jitter()+labs(x = expression(Sulphates~(g/dm^{3})) ,
                   y = expression(Fixed~Acidity~(g/dm^{3})),
      title = "Relationship between Fixed Acidity ,Sulphates & pH")+theme_bw()+
scale_color_gradient(low = "green", high = "red")
```

* From above plot we can deduct that as sulphate value increased there is a 
slight increase in pH value
* Also, when fixed.acidity is increased there is a decrease in pH value
* There seems to be negative association between fixed.acidity and pH
* Majority samples have fixed acidity above 6 and sulphates less than 0.6


**Relationship between Alcohol , Residual sugar and Quality**

```{r echo=FALSE,message=FALSE, warning=FALSE,alcohol_sugar_and_quality}
#Scatter plot for Alcohol , Residual sugar and Quality
ggplot(whiteWineData,aes(x = alcohol, y = residual.sugar ,
                         color= as.factor(quality)))+
geom_point(alpha=0.5, size = 1, shape = 1)+ 
geom_jitter()+labs(x = "Alcohol (% by Volume)",
                   y = expression(Residual~Sugar~(g/dm^{3})))+theme_bw()+
scale_color_brewer(type = "div", palette = "RdYlGn", name="Quality", 
direction=1) +guides(color=guide_legend(override.aes = list(alpha = 1)))
```

* From above plot we can confirm that alcohol % also contributes to the quality.
* Samples with alcohol % above 11 have higher levels of quality and vice versa
* Majority samples have residual.sugar level below 10 
* AS residual.sugar increased higher quality levels are decreased  which 
suggests residual.sugar effects quality to some extent
 
# Multivariate Analysis

### Talk about some of the relationships you observed in this part of the \
investigation. Were there features that strengthened each other in terms of \
looking at your feature(s) of interest?

* Majority samples have fixed.acidity above 6 .
* Distribution of pH values w.r.t fixed.acidity & sulfates is normal and
 there are some outliers. This is further strengthened by Bivariate analysis 
 between pH & sulphates , pH & fixed.acidity

### Were there any interesting or surprising interactions between features?

* There is inverse correlation between fixed.acidity and pH value
* Higher quality samples are seen when alcohol is above 11%

# Final Plots and Summary

Below are 3 plots with most interesting findings

### Plot One
```{r echo=FALSE, Plot_One}
#Plot one : Bar plot for Ideal pH
ggplot(data=whiteWineData, 
  aes(x=idealpH,fill=idealpH)) + 
geom_bar(aes(y = (..count..)*100/sum(..count..)))+
labs(title=paste("Bar plot for Ideal pH"), x="Ideal pH(Yes/No)", y="Total Percentage(%)")

```

### Description One

 Ideal pH range of white wine is in between 3-3.4. From above plot we can see 
that more than 80% samples are in ideal pH range.

### Plot Two
```{r echo=FALSE, Plot_Two}
#Plot two : Scatter plot of Ideal pH and Quality
ggplot(whiteWineData,aes(x = idealpH,y = as.factor(quality),color=idealpH))+
geom_point()+geom_jitter()+labs(x = "Ideal pH value for white wine(Yes/No)",
y = "Quality(0-10)",title="Ideal pH vs Quality")
```

### Description Two

* From above plot we can confirm that majority samples with ideal pH range have
 quality levels in between 5,6 and 7 compared to samples not having ideal pH
* quality has some positive association with ideal Ph value, which strengthens 
that ideal pH range is 3-3.4

### Plot Three
```{r echo=FALSE, Plot_Three}
#Plot Three: Scatter plot for Alcohol , Residual sugar and Quality
ggplot(whiteWineData,aes(x = alcohol, y = residual.sugar ,
                         color= as.factor(quality)))+
geom_point(alpha=0.5, size = 1, shape = 1)+ 
geom_jitter()+labs(x = "Alcohol (% by Volume)",
                   y = expression(Residual~Sugar~(g/dm^{3})))+theme_bw()+
scale_color_brewer(type = "div", palette = "RdYlGn", name="Quality", 
direction=1) +guides(color=guide_legend(override.aes = list(alpha = 1)))
```

### Description Three
 
* From above plot we can confirm that alcohol % also contributes to the quality.
* Samples with alcohol % above 11 have higher levels of quality and vice versa
* Majority samples have residual.sugar level below 10 
* AS residual.sugar increased higher quality levels are decreased  which 
suggests residual.sugar effects quality to some extent
 
# Reflection

* This is the tidiest dataset and it was easy to perform Univariate analysis.
* For Bivariate Analysis I couldnt figure out main attributes and supporting 
attributes initially, resulting in some re-work.
* After some research on white wine I was able to determine.Which suggests that 
prior knowledge of dataset attributes is required to make a solid analysis.
* Some of the important attributes of wine like age, tannins,types of grapes etc
are not mentioned in dataset which would have helped in understanding more about
quality.
* Given this Data would help in understanding more about wine and finding other
main attributes that contributes to quality.

# References
 * https://winefolly.com/review/wine-characteristics/
 * https://winefolly.com/tutorial/wine-sweetness-chart/
 * https://www.winespectator.com/drvinny/show/id/5035