Simple Analytics Approach - Residual Analysis (TS E17)

In the last video we created a simple approach of using the daily average to predict the same calendar day moving forward. In this video we review a few error metrics and plot out the errors. It is crucial to really review your errors as they will tell you what you have done correct and what you have done wrong. Do not be a fool and try to minimize your errors. The model structure and logic used to build the model should be the main focus however your errors tell you if you were close to correct and what you missed. From these errors it is clear that the simple average by day approach has serial correlation and heteroskedasticity. This means we have not fully addressed the complex structure of our data. We will move on to statistical approaches and neural networks to see if we can directly model the serial correlation and heteroskedasticity. We also cover histograms, normal distributions, and kernel density estimation in regards to residuals (errors). Listen to your data! Understanding the Different Model Error Calculations: https://youtu.be/_QmEkX8Q41o FULL TIME-SERIES PLAYLIST: https://www.youtube.com/playlist?list=PLBfqPS8Xvt2D2pagOjSEkQYCcn_8X-vzg SUPPORT THE CHANNEL: Quant t-shirts, mugs, and hoodies: https://teespring.com/stores/fancy-quant Connect with me: https://www.linkedin.com/in/dimitri-bianco https://twitter.com/DimitriBianco ☕ Show Your SUPPORT and Buy Me a COFFEE ☕ https://ko-fi.com/fancyquant