Lab Assignment 02: Exploring Palmer Penguins Dataset

Step 1. Understanding Penguin dataset (15pts)

The first step of exploratory data analysis is to explore the number of instances, number of features, and type of each feature

Download palmer_penguin.csv data from the course schedule

Create a new colab notebook. Make sure you add your name

1. Import pandas and matplotlib libraries

2. Import data and name it dataset

3. Display the first 5 columns

4. Identify the shape of the dataset

5. Identify the information about features types

Step 2 Exploring Relationships (20 Points)

Once the dataset's size and features are understood, we can explore the relationships between features. Data visualizations like scatter plots give a visual representation of how pairs of features change together. Instead of creating individual scatteplots, you will learn how to use pair plot with seaborn library.

1. Import seaborn library

import seaborn as sns

2. Seaborn's pairplot() creates a grid of pairwise scatterplots for numerical features in the penguins dataset:

p = sns.pairplot(dataset)
plt.show()

3. Increase the font for all future plots, using update(). Then rerun your pairplot:

plt.rcParams.update({'font.size': 14})

4. Add categorical values to the plot. The hue parameter adds color based on the values of the island column:

p = sns.pairplot(dataset, hue='island')
plt.show()

5. Replace island by species category
6. Create a contingency table (=crosstable) for species and island

pd.crosstab(index=dataset['species'], columns=dataset['island'])

7. Create a barplot with species but instead of using barplot we will use a simplified version with seaborn
   • Take a look at https://seaborn.pydata.org/generated/seaborn.countplot.html
   • Create a countplot for species
   • Create a countplot for species but color-coded with island

Step 3 Describe the shape of data (15 Points)

The shape of a feature's distribution is important way to examine if data is normally distributed, or skewed

1. Calculate descriptive statistics using describe()
2. Take a look at the documentation https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.describe.html and include all data. The default is only for numerical
3. Create histograms for numeric data (should be 4). See how to drop the last column year:
To drop year column dataset.drop(columns=["year"]).hist()
4. Adjust the figure size to fit 4 histograms better. Then rerun your plot

plt.rcParams["figure.figsize"] = (12, 10)

5. Change the size of the bins, try different numbers. Check out how to change bins: https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.hist.html
6. Add title (Come up with a good title)

Step 4 Detect missing data (20pts)

If only a few instances have missing data, those instances may be set aside without having a negative impact on a model. if many instances have missing data, techniques like imputation may be needed

1. Use pd.isna() to identify missing values for bill length

dataset[pd.isna(dataset['bill_length_mm'])]

2. Display missing data counts for all columns (see In-Class Labs)
3. Import missingno library - see In-Class Practice
4. Use bar() amd matrix() to visualize data

Step 5 Impute missing data (20 Points)

1. Impute bill_length_mm column with mean()

#Follow this format:
dataset['column'] = dataset['column'].fillna(dataset['column'].name()) 
# where name is the imputation method
    

2. Impute bill_depth_mm column with median()
3. Impute flipper_length_mm column with your own choice of method
4. Impute categorical missing data with the word "Unknown". HINT: it will be just the string value
5. Make sure all missing data is imputed. Run info()
6. Finally, display info() with dropped na dataset