If you could travel to the future from when you posted this video, you would come to a conclusion that in 2024 this is still extremely relevant. Great presentation!
It was funny to see that the host and the presenter wore the exact same outfit. It took me by surprise when the host started speaking at the end of the talk (zoomed out view) and the voice was very different. Great talk and awesome answers Kay !
You, sir, are an outstanding presenter. This was perfect. Thank you for developing and releasing the package and thank you for providing this excellent presentation.
Wow this video was posted 7 years ago and it is still so relevant today, in the world of “AI”. Great presentation and some of the questions answered at the end were spot on. Thanks for the awesome presentation!
28:21 "can we use this to measure multiple events that overlap in time?" "that's an open research question" Well I hope someone answered that in the last 8 years because that's why I'm here
Thank you for the presentation. Must be noted that the validity of this approach is entirely dependent on the ability to accurately extrapolate the control scenario based on observational data. In a nutshell, it relies on our ability to 'randomize' the treatment and control groups via adequate control variables. If, on the other hand, we miss a key variable in our extrapolation model, the estimated causal effect of the variable in question will be biased. This causal estimation is nothing but a way to approximate a randomized experiment scenario via a model which attempts to control for all relevant outcome drivers.
окей, если несколько тритментов, почему мы не можем посчитать один, а потом его вычесть из временного ряда во втором тритменте, чтобы оставить влияние только одного изменения?
If you could travel to the future from when you posted this video, you would come to a conclusion that in 2024 this is still extremely relevant. Great presentation!
It was funny to see that the host and the presenter wore the exact same outfit. It took me by surprise when the host started speaking at the end of the talk (zoomed out view) and the voice was very different.
Great talk and awesome answers Kay !
You, sir, are an outstanding presenter. This was perfect. Thank you for developing and releasing the package and thank you for providing this excellent presentation.
Wow this video was posted 7 years ago and it is still so relevant today, in the world of “AI”. Great presentation and some of the questions answered at the end were spot on. Thanks for the awesome presentation!
28:21
"can we use this to measure multiple events that overlap in time?"
"that's an open research question"
Well I hope someone answered that in the last 8 years because that's why I'm here
One of the best talk I've ever seen this is how you should explain things.
Sure
Kay has very concisely explained what a super power a Marketing Analyst has to their exposure. He is a really good presenter.
I was reading the reaseach paper but got to know about your video and my week work covered by your 30:38 mins. Amazing presentation!!!!
One of the best presentations. Thanks a ton for explaining this so beautifully!
Its all imputation finally :) I loved the approach! Thanks for the presentation!
Thanks, Kay Brodersen
Rajavel KS, Bengaluru.
He is amazing presenter! Many thanks
great talk and great questions from Q&A session. want to know more about how to choose the predictor time series.
FANTASTIC Presentation!
i would like to make the option at 16:30 - pre-intervention, intervention and post-intervention, not only the classic pre and post period
What if we dont have a high correlated time series to train the model ¿
Excellent presentation.. Thank you!
Thank you for the presentation. Must be noted that the validity of this approach is entirely dependent on the ability to accurately extrapolate the control scenario based on observational data. In a nutshell, it relies on our ability to 'randomize' the treatment and control groups via adequate control variables. If, on the other hand, we miss a key variable in our extrapolation model, the estimated causal effect of the variable in question will be biased. This causal estimation is nothing but a way to approximate a randomized experiment scenario via a model which attempts to control for all relevant outcome drivers.
Exactly I agree with you, It is like a prediction based on a prediction.
Wow! Really cool library, amazing presentation. Can't believe I didn't know this was a thing.
Excellent Presentation on Causal Analysis
Thank you for this video, it helped me a lot for my research. you're great
Amazing presentation, I am currently working with this tool and you made things so much clearer, congratulations
A very clear presentation about the CAUSAL IMPACT tool! Thanks!
Can independent variables in the above example be considered to be instrumental variables?
окей, если несколько тритментов, почему мы не можем посчитать один, а потом его вычесть из временного ряда во втором тритменте, чтобы оставить влияние только одного изменения?
I keep being amazed!
This made my career
what if we only got post period, is it feasible to do it?
Such a great explaination! Thank you
Fantastic talk. Thanks for sharing
Great video
Great talk!
Thanks! I like the summary() function!
This is so amazing
excellent explanation!
❤
How is CasualImpact different from a common marketing mix modelling project?
Fantastic talk! Thanks for sharing