Module 1-5 - Data Analysis

Question 1

Ways to deal with missing data? (3)

Answer 1

1) Replace missing values with valid values that don’t bias the data (mean, median, mode)
- Assuming that missing values are missing completely at random

2) Replace missing values with an assigned category/value
- Assign a special “code” for the missing values

3) Predict the missing value using the other variables

Question 2

Random sampling, description?

Answer 2

Draw random # of records (without replacement) from the dataset until you have the required #

Question 3

Stratified sampling, description?

Answer 3

Independently drawing a set of random records from each strata in your data
-Oversampling and undersampling

Question 4

Systematic sampling, description?

Answer 4

Drawing samples according to a pattern (ex: every 5th record)

Question 5

Univariate data exploration, checklist? (4)

Answer 5

1) Understand basic relationships in data
- Common sense checks on model output

2) Identify potential data errors that could cause misleading models

3) Identify outliers
- Understand their potential effects on the model

4) Understand how target/response varies for different predictors

Question 6

Ways of gaining insight into variable distributions? (2)

Answer 6

1) Calculate numeric statistics/summaries of the data

2) Visualizing values of a variable in a graphical image

Question 7

Types of variable combinations we can consider when examining bivariate relationships? (3)

Answer 7

1) Categorical vs Categorical
2) Categorical vs numeric
3) Numeric vs Numeric

Question 8

Why are outliers bad? (3)

Answer 8

1) Can hinder model’s ability to find the patterns in the data to predict an outcome
2) Model will try to optimize its predictive accuracy by minimizing prediction error
3) If too many outliers, model will bias its predictions toward the outliers in order to bring down the overall error of the prediction

Question 9

Visualization of bivariate errors, which type of plot to use:

1) Numeric vs Numeric
2) Categorical vs numeric
3) Categorical vs Categorical

Answer 9

1) Scatterplot
2) Box Plot
3) Frequency Table

Question 10

Checklist for when you do univariate data exploration?

Answer 10

1. Range of the variable, different factor levels
2. Skewness: mean vs median (numeric)
3. Any transformations to do with it? Any new variables to create?
4. Show the graphs

Question 11

Visualization of bivariate errors, which type of plot to use:

1) Numeric vs Numeric

Answer 11

Scatterplot

Question 12

Visualization of bivariate errors, which type of plot to use:

2) Categorical vs numeric

Answer 12

Boxplot

Question 13

Visualization of bivariate errors, which type of plot to use:

3) Categorical vs Categorical

Answer 13

3) Frequency Table