?? Day15 of #100DaysOfPython ??

Today, we're diving into different types of techniques for handling missing values in a dataset!

Checkout the what is missing data and its different types in Day14 Article.

Q. What are the different types of techniques for handling missing data?

1. Imputing missing values with mean/median/mode
2. Random sample imputation
3. Imputing NaN values with a new feature
4. End of distribution imputation
5. Arbitrary imputation
6. Frequent categories imputation

In today's article we will start with -

1. Imputing missing values with mean/median/mode:

• This imputation works on the assumption that the data is missing completely at random (MCAR).
• In this method we replace the NaN with the most frequently occurring observation in the feature.

Let's dive deeper with an example below:

Link to titanic dataset

Imputing feature with mean/median/mode can have the following:

Advantages:

• Robust to outliers
• Faster way to obtain complete dataset

Disadvantages:

• Change/Distortion in the original variance of the data
• Impacts correlation

What real world examples can you think of where imputing missing data in feature with mean/median/mode would be a better choice?