Yesterday’s example made me think a lot. Below is equation for a correlation coefficient between two data sets:
At first I thought that in very simple words if correlation equals 1 then trend lines are equal. Let’s have a quick look on a very simple example:
This two primitive data sets are having very similar trend lines, however their correlation is only 0,4. Another example:
Numbers in two data sets are scaled. Correlation equals 1.
And a final example:
Very high correlation at the level of 0,91. Interesting.
I do understand why in yesterday’s example it didnt make sense to correlate daily changes, which would end up in completly irrelevant and non correlated trend lines.
It’s time to dig more into time series forecasting. Quick google resulted in a blog post I want to analyze in comming days: https://towardsdatascience.com/bitcoin-price-prediction-using-time-series-forecasting-9f468f7174d3 and this one: https://machinelearningmastery.com/arima-for-time-series-forecasting-with-python/
Thanks for reading,