I started learning about time series by myself as an adult about 4 years ago. Bought books from amazon but never truly understand the concept behind it, most books just list a bunch of formula. But after watching many of your videos, I realised it is the way of teaching that make things difficult. Sir, you are an exceptional teacher. I truly believe you will excel in all of your future endeavours. I hope the channel will keep running for as long as possible, and many others including myself will be able to learn from you. Thank you so much!
Many thanks for uploading these videos. They are outstanding! I really enjoyed all of your videos so far and I cannot wait to see more of your videos. I would love to see something about ARMAX (and the insertion of exogenous variables into the models in general) and maybe something more about the probabilistic/confidence of the predictions.
Hi, just wanted to say how useful your videos are. really appreciate its both practical and has a good level of intuition/theory. looking forward to your next uploads!
We need to recognize that some authors prioritize fame over ensuring their books are easily understood by readers. For them, the primary goal may be gaining recognition rather than ensuring their content is accessible and comprehensible to their audience.
Thank you so much for the content, they really help us understand these concepts better. It would be even more helpful if you could please provide links to the code in the description.
Wow 50k! Well done. I remember when you had 8k followes and I was wondering why doesn't this guy get more followes?! On the right track mate. I'll write another comment when you hit 500k :)
thanks for the vido Once the PACF graph is drawn and we see the lags we can consider for the AR model, what code do we use to get our model done only with those lags and not those within the error zone???
How do you "exclude" certain lags from the tsa ARMA function - does it have a parameter for that or something? For example you determine lags 2, 4 & 5 are insignificant yet when you call ARMA(7,0) it still does all 7. Thanks
Following up - how do we exclude particular lags from the model? At around 5:50 you determine lag 2 is not significant yet can we explicitly exclude that?
Lag 2 is not a good predictor since p-value is high. Will we just drop lag 2 and proceed with the rest of the equation as the final equation? I thought we need to re-run on just lag 1 and 3 and it will change the coefficient of the equation and then only we can use that equation. If yes, could you please tell how to run AR(3) Model and not include lag 2?
@@tomislavprimorac1050 To not include lag2 in the model, you'd create lag 1 and lag 3 variables manually (by shifting the original data). Then we use an ARIMAX model where p=0, q=0 and the exogenous variables are the lag1 and lag3 variables you created
Got a simpler way to make train and test data: train_end = int(len(df)*0.7) train_data = df.iloc[:train_end] test_data = df.iloc[train_end:] It splits the dataset into two sets, each 70% of the dataset and 30%
You are dealing with timeseries data with a nicely typed datetime column. In this context showing intent (which specific date you are considering for the test) I would argue is much clearer.
HELLO SIR I AM A FREELANCE ENGINEER WHO MAKES MONEY BY MAKING PROJECTS BASED ON CLIENT NEEDS I WOULD LIKE TO ASK YOUR PERMISSION TO USE YOUR CONTENT AND TO USE YOUR CODE TO LEARN AND MAKE DIFFERENT PROJECTS.I WONT BE SELLING YOU CONTENT TO ANY WEBSITE