All models are wrong?
This is one of the boldest sentences I have heard over the past years.... and you know what? I definitely agree with it!!!
Attending a Data Science training our teacher made reference to this known sentence from George Edward while making us understand that wasting lots of energies trying to optimize analytical models might not make sense. This is due to no matter how far we arrive our model will always be scientifically wrong due to statistical assumptions we are making while making it up.
From my professional experience I can't remember one single time where one of hardest choices to be made is determine whether the model and conclusions ellaborated were enough to carry on or maybe another twist should be required... isn't that your case generally?
Looking for an analogy of stock markets having a "stop loss" strategy is something to be naturally adopted by data analysts... it basically means assuming that our model will never be perfect, stablish a specific amount of resources to develop it with the commitment to carry on conclusions once that milestone is achieved. Our business shouldn't be impacted by "Paralysis by analysis" behaviours since that's probably the most harmful scenario far beyond from making choices with a certain level of uncertainity.
Can your share your comments of experiences on this I would love to hear!
Attending a Data Science training our teacher made reference to this known sentence from George Edward while making us understand that wasting lots of energies trying to optimize analytical models might not make sense. This is due to no matter how far we arrive our model will always be scientifically wrong due to statistical assumptions we are making while making it up.
From my professional experience I can't remember one single time where one of hardest choices to be made is determine whether the model and conclusions ellaborated were enough to carry on or maybe another twist should be required... isn't that your case generally?
Looking for an analogy of stock markets having a "stop loss" strategy is something to be naturally adopted by data analysts... it basically means assuming that our model will never be perfect, stablish a specific amount of resources to develop it with the commitment to carry on conclusions once that milestone is achieved. Our business shouldn't be impacted by "Paralysis by analysis" behaviours since that's probably the most harmful scenario far beyond from making choices with a certain level of uncertainity.
Can your share your comments of experiences on this I would love to hear!
Comments
Post a Comment