Selection bias

Conquering Data Bias

Konstantinos Kattidis

Data Analytics Lead

What is selection bias?

It's the bias introduced when the data for analysis is selected in a way that systematically favors certain individuals, groups, or characteristics

Diagram showing the population that is different from the sample.

Hence, the sample obtained is not representative of the population intended to be analyzed

Let's delve into the five common types of selection bias

1. Sampling bias

Sampling bias occurs when the sampling method is not fair or random
It originates from the approach we choose to obtain our sample which can make it hard or impossible to apply the findings to the whole population
For example:
- An e-commerce platform analyzes customer satisfaction using convenience sampling
- The findings may not reflect the sentiments of the entire customer base

Person thinking which sampling method to use

2. Undercoverage bias

Consider a market research study targeting online consumers, excluding individuals without internet access
Undercoverage bias highlights the inadequate representation of certain groups within the chosen sample
It is distinguished from sampling bias by its focus on the representation of specific groups rather than the randomness or fairness of the sampling method itself

Researchers ignoring consumers without internet access

3. Non-response bias

Non-response bias arises when individuals who choose not to participate in a survey or study differ systematically from those who do participate

Satisfaction survey where respondents are satisfied and non-respondents are dissatisfied

In a survey assessing employee satisfaction dissatisfied employees are less likely to participate
This leads to to an overly optimistic view of employee morale

4. Self-selection bias

Analyst exploring results affected from self selection

Self-selection bias occurs when individuals choose to participate in a study or provide feedback
For example:
- Customers self-select to participate in a satisfaction survey
- When their views do not represent the broader customer base
- This skews the overall perception

5. Survivorship bias

It occurs when only successful entities are included in the analysis

For example:

Analyzing successful product launches without considering the ones that failed
This would lead to biased insights, overlooking critical factors that contribute to failure

Person analyzing successful products and ignoring the failed ones

Creating a cohesive understanding

Scale unbalanced with question mark

It's not uncommon for multiple biases to interact, complicating analyses
For example, a customer satisfaction survey may exhibit both self-selection bias and non-response bias

Let's practice!

Conquering Data Bias