1. Introduction to key influencers
Welcome back for the final lesson! We spent the past several chapters learning how to familiarize ourselves with variables, visualize distributions, quantify their characteristics with statistics, and explore relationships.
In this final part, we will use another Power BI visualization to identify key influencers affecting the change of a target variable.
2. What are "key influencers"?
What are "key influencers"?
They are explanatory variables which cause a significant change - increase or decrease - in the statistics of a target variable.
The Key Influencers visualization is helpful when looking to determine major drivers in a metric, comparing the relative importance of these drivers, and simply exploration of relationships between a target and another variables.
3. What are "key influencers"?
You explored the foundations of this concept starting in the second chapter when evaluating relationships between two variables.
Remember finding that the average monthly income changes depending on the stage of career a professional is in?
As seen by the box plots, being in a "tenured" career stage may be considered a key influencer on the average monthly income of professionals as the median average monthly income is visually (and numerically) larger than the other two groups.
4. What are "key influencers"?
Similarly, you discovered that average monthly income is also influenced by job level and job role.
This scatter plot, created in an earlier exercise, shows that a job level equal to 5 results in an average monthly income of about $19,000 - almost $15,000 more than the total dataset sample average.
Likewise, Sales Representatives make $2,300 on average which is just under half of the total dataset sample average.
5. Key influencer visualization
The Key Influencer visualization, in a way, expedites the EDA process. Specifically, it leverages regression models behind the scenes to identify important variables. This is beyond the scope of the course.
Here is an example created to find influential variables leading to an increase in monthly income. Job role, total working years, years at company, years in role, and years since last promotion were used as the "explanatory" variables.
6. Key influencer visualization
Zooming in, we see the job roles Manager and Research Director were marked as key influencers. When these titles were present, the average monthly income increased by $11,760 and $10,300, respectively.
Power BI doesn't only identify variables but more specifically the values which are important in the changes observed in the target variable.
7. Reasons why key influencers are not found
Key influencers will not be found within each variable added to the visualization. Sometimes, an influence won't be found at all. This could occur for a couple of reasons.
First, the explanatory variable may have too many categories. This can lead to inability to generalize an observed pattern to one of the many categories. If this is the case, it is often helpful to transform into a new categorical variable with a consolidated number of categories.
Second, there are not enough observations to derive patterns. A large enough sample size is necessary for any meaningful analysis. If too small, insights are limited or elusive altogether.
Lastly, your explanatory variables may have enough observations to generalize, but the visualization simply didn't find any meaningful correlations to report. In this case, it's helpful to explore other explanatory variables.
8. Let's practice!