Видео

I really appreciate your lucid explanations in all your presentations on this subject. In this video you did touch upon the of mean and standard deviation values for the sampled conditional distributions , asking the viewer to see earlier videos for explanation. Perhaps it would improve the presentation, if you at least mentioned that it is due to the bivariate dustribution's correlation parameter (rho = 0.5).

3 years later and still the goat

Really helpful~ Just got stuck at line 2 of the derivation. thought L(1) == C(t-1) ?

watching your 2nd video. Great explanation! The best thing is intuitive understanding. Thank you for help in learning)

I love this guy😭

Did you do training-test split on the data?

I'm a simple man. When Ritvik posts, I watch.

Knocking it out of the park, as usual.

Great video . Can you do some Multivariate analysis data series with different business case scenarios

What an excellent teacher you are! Thank you for all your videos.

Great Explanation

How do you decide whether to accept a sample proposed by g(x)? For example, say f(s) / (M*g(s)) = 0.4, do you accept it or not ?

Amazing

I would add that it's not experimental but quasi experimental because it's not really granting that the effect is due to our treatment. It might be due to a common cause or be causally biased by it so we are also assuming causal sufficiency and a sufficient knowledge of the underlying process. Still we have no way to guarantee that some unobserved cause happened in that time window. Good and clear video, as usual btw, keep going bro °u°

gender pay gap is fake

Great content! Thanks for posting it

Stationarity conditions - mean = c - var = c - no seasonality or periodic repetiotions

ACF - takes into account all possible t-k which can affect t. Calculated using Pearson coefficient. Direct and Indirect effects. PACF - takes into account only t-k, and how it affects t. A regression line is fitted for this. The coefficients obtained explains the affect of t-k on t. Direct effects only.

Dear ritvik why does the determinant value of the first explanation results comes out to be zero(I did try to change some conditions like a single path to every next point but then also that value also comes out to be zero ) what does it indicates?

Notes for future - TS is an extrapolation problem and error keeps on increasing as we move away from known data - Reg is an interpolation problem and error is more or less same, since prediction is usually made in the range of available data.

saved my life again!

I am still confused about how you developed the kernels in the first place. I know what they do but don't know how to obtain them without using the transformed space.

hi R, one ques. in what sequence should we watch these lectures ? the playlist seems to be jumbled , or it serves a purpose ?

Hi Ritvik! The way you've set up the linear model(s), the intercept parameters B0 and B2 will represent the intercept from Tt=0, rather than at the particular time that the interruption happens. So, if you got a large positive change in the B3 slope parameter, you'll probably get a negative change in the B2 parameter (since a steeper line at positive Tt will intercept the yt axis *below* where the original line did! Wouldn't it make more sense to do a shift/translation of the Tt parameter so that it works as if the interruption happens at the yt axis? For example, use a translated variable Ut = (Tt - t). So, if the interruption happens at t = 75, then when Tt = 75, Ut = (Tt - t) = (75 - 75) = 0. And your model can then be: yt = B0 + B1⋅Ut + Dt⋅(B2 + B3⋅Ut) At least this way, the B0 and B2 params will have a much more interpretable meaning. B0 will be the value of yt at Tt = t of the main linear model, and B2 will be the *initial change* in vt at Tt = t. In other words, how much 'immediate effect' did the interruption have; how much of a vertical 'jump' up or down. Granted, to calculate vt at any particular Tt, you'll have to first convert to Ut, but that's not so bad, just a simple shift. And if you really need to find the linear params in terms of Tt, it's fairly easy to just plug in Ut = (Tt - t) and expand out to find the transformed linear params for Tt.

Thank you so much

How can we take advantage of the GPU?

Thanks for this tutorial on ITS. I have to admit, I am a bit confused by this example. I wasn't sure at first why. But basically, it is because the "data" is "too simple". At first, I look at this and I say "Any one" can see that Pre-advertising is not as good as "post" advertising treatment. So, Why would I need to use this technique. Why not just count sales per "month" and say, yup ... we boosted total sales per month. There's nothing inherent in this data sample that confounds the analysis or makes it difficult or non-obvious. On the other hand, if you could show something where the sales data points are more difficult to track, so that it is non-obvious by immediately looking at the data (in graph form) that the trend after treatment is looking seemingly more positive, but using ITS we can see that while visually it seems to look better, in fact, it is not as good after treatment (per this method.) And perhaps a second example, where it seemingly looks more negative after treatment, but by using ITS we can prove that it is better after treatment Then, I think those would give more credence to this method. Because as far as I can tell, the graph makes it far to easy too see the trend... so why bother with ITS unless it can show something that isn't obvious. And, again, thanks for this explanation on Interrupted Time Series.

i still dont get the part of getting the effect without the experiment.... if i dont have the time series data of my ice cream shop where i decided to advertise as right now im in middle of taking this decision then how can i fit a model which can measure the effects of it?

I confess, I didn't understand any single word of explanations from my MIT professor about perceptron. How ever, after I saw this video and understood clearly what the idea is. Thanks.

thanks, may I ask what do you do for a living?

the equation of the hyperplane is w·x+b=0, isn't it? The video says - b=0 instead of +b=0

Great vid. Please explain how to compute the confidence intervals.

Absolute legend

Man your channel would blow up spectacularly if you invested the time into learning how to make really nice visuals.. the whole poorly hand drawn example thing is really 2005 && screams laziness and/or amateur..

Excellent video and content, thank you. I am not a trader nor a financial person but the training window size vs. the holding window size may be part of the problem unless the choosen entry time would be an outlier by chance. The holding size may play a huge role in certain market dinamics with periods of revovery much greater than the chose holding window time. When market goes down correlations tend to one and in many cases do revert to positive from negative in calm situation. Best

I thought U is mxm, V is nxn and SIGMA is mxn

Can you show us the code for this, especially for VAR

This means that apples are far better than bananas right? So when I've been told that you cant compare apples and bananas ... thats a lie ...

Thank you for the video. it was nicely explained. There are a lot of simplifications. Could you also talk about how best select sigma and l - is it all done empirically? also do you have any example of implementation?

Pls recommend a book for data science n machine learning

Very clear and helpful. Thank you so much!

So simple… but so powerful!

Go on, tell us how to test if and when a change occurred :)

Any data science text book recommendations?

Thank you for this

I would argue that 2 has seasonality...

Hmmm i noticed that if two categories are strongly correlated, the plot will look close to a straight line. Going to multidimensional space, that "line" looks like the vector u1 in the video, on which the data are projected. Does that mean PCA will perform better the more correlated two (or more) categories are?

Can you do this again? Polls mean nothing. There are too many confounding variables.

amazingly taught. thank you so much

nice