Machine learning in trading: theory, models, practice and algo-trading - page 3131
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Can all this wonder content be put in a separate thread?
I'm done
Can you show me something like that?
If it's that common.
Finally!
Well, there's a start.
What's next?
What's so difficult about it? Jumping between scales is something we've done before, there are no secrets here. I'm looking for more complicated things, and a picture with just one deal does not reveal the essence of the whole picture in any way.
I see.
It's not hard to write letters.
I get it.
It's not hard to spell, of course.
Have you seen my freelance work?))))
Have you seen my freelance work?))))
What methodology can be used to assess the quality of deals when seeing a freelance profile?
Randomisation removes the bias between test and control, after which the predictor impact is estimated
If the bias is not removed beforehand, it will be associative rather than causal.
The gold standard
In the previous lesson, we looked at why and how association differs from causation. We also saw what is required for an association to become a causal relation.
E|Y|T = 1] - E[Y|T = 0] = E[Y - Yo|T = 1] + {E[Yo|T = 1] - E[Yo|T = 0]} ATT ADJUSTMENT.
Recall that association becomes causation if there is no bias. There will be no bias if E[Yo T = 0] = E[Yo T = 1]. In other words, the association will be causal if the treated and control patients are equal or comparable except for their treatment. Or, to put it in more technical terms
Above is a translation of the picture.
To begin with - I can't understand at what point you want to split the sample into two subsamples.
Next - apparently there is a special terminology here, causation is a direct effect on an outcome - perhaps not even a probabilistic pattern anymore. An associative relationship is either an activator of the cause or an associated feature, and is usually probabilistic.
I don't understand the formula - state the point in human terms?
But, the point of these methods (UpLift) is to estimate the factor that exclusively influenced the target. I understand that the degree of influence is assessed. And, let's say, in our case we don't know such a factor and we go through everything - we get some measurements as an output. And what do you suggest we do with them? Exclude bad indicators?
How can we use this with gradual drift of data?
I don't exclude it, maybe you have come up with something brilliant, but I haven't caught the train of thought yet.
The gold standard
In the previous lesson, we looked at why and how association differs from causation. We also saw what is required for an association to be a causal relationship.
E|Y|T = 1] - E[Y|T = 0] = E[Y - Yo|T = 1] + {E[Yo|T = 1] - E[Yo|T = 0]} ATT ADJUSTMENT.
Recall that association becomes causation if there is no bias. There will be no bias if E[Yo T = 0] = E[Yo T = 1]. In other words, the association will be causal if the treated and control patients are equal or comparable except for their treatment. Or, to put it in more technical terms
Above is a translation of the picture.
For starters - I can't understand at what point you want to split the sample into two subsamples.
Next - apparently there is a special terminology here, causation is a direct influence on the result - perhaps not even a probabilistic pattern anymore. An associative relationship is either an activator of the cause or an associated feature, and is usually of probabilistic significance.
I don't understand the formula - can you give me the gist of it in human terms?
But, the point of these methods (UpLift) is to estimate the factor that exclusively influenced the target. I understand that the degree of influence is assessed. And, let's say, in our case we don't know such a factor and we go through everything - we get some measurements as an output. And what do you suggest we do with them? Exclude bad indicators?
How to use it in case of gradual drift of data?
I don't exclude, maybe you have come up with something ingenious, but I haven't caught the train of thought yet.
you can ask chatgpt for formula decoding if you don't understand any of the symbols.
Y|T = 1 test group outcomes (with tritment)
Y|T = 0 - control group (without)
Y - class label, Y0,Y1 - class labels without and with the tritment.
T - tritment introduced into the model (including predictor) or not introduced (1;0)
E - expectation
Split at any point as you divide by test and traine
If you don't do mixing, you get a biased estimate of ATE+bias
ATE is the average treatment effect of the exposure
sleepy, I may mix up the letters in places, but the logic should be clear
by the way, google's bard is more to my liking than gpt. It can google and it's free.
but it only supports English and vpn in the US or England, it doesn't work in other countries.
And basically, who are openAI and who are the Googles. Probably different weight categories.by the way, google's bard is more to my liking than gpt. It can google and it's free.
but it only supports English and vpn in the US or England, it doesn't work in other countries.
And basically, who are openAI and who are the Googles. Probably different weight categories.