Why a Strong Correlation Isn't Necessarily a Good Thing | Statistics for Marketers Series скачать в хорошем качестве

Why a Strong Correlation Isn't Necessarily a Good Thing | Statistics for Marketers Series

Performance Marketers fall prey to just relying on correlation all too often. A classic example, the one I also discuss in the video, is when trying to understand the relationship between ad spend and sales/revenue. The most common mistake is to stop at correlation analysis alone. But a high correlation alone doesn't say anything - if multicollinearity is present among your independent predictors, your correlation analysis will be misleading. Similarly, a low correlation coefficient doesn't necessarily mean there is no strong association - I show with the example of a non-linear relationship being modeled by a linear regression, which fails to capture the true nature of the relationship between target and input variable. Finally, I also show the case of interaction terms that also can reveal the importance of a variable in predicting the outcome (Eg: sales) even if individual correlation is small. This is the first part of a series I will be looking to launch in the coming days to help marketers sharpen their understanding of statistical principles. #marketinganalytics #performancemarketing #marketingmeasurement #statistics #datascience