1. Is sql mandatory for product management? Or it’s a plus. Can you tell me more about the technical skills required for product management? Thanks

  2. Without it you’re just a manager who has to ask analyst to pull some data every day

  3. There’s a huge spectrum of jobs involving psy+ds knowledge. From modeling to consulting. I would advise you determining a bit better what are you interested in, because after some psystat classes there is no coming back to plain ml algos (eg phd in hardcore psychometrics)

  4. You can impress someone on an interview for a psychometric position with dif?

  5. Basically it’s the effect of external factors influence on the decision. How do you negate this? Be an expert, know what you are doing, and have the ability to communicate it.

  6. Actually know what is a right thing to do

  7. IMO it is better to state that this represents k-fold validation (which in itself could lead to errors), and that the bias-variance tradeoff is useful, but not universal

  8. I believe it’s from a Twitter thread

  9. Yes, between YLecunn and some other guy

  10. Most of the time the PM interview process is a joke and both sides agree with it and just play along

  11. I would say it require quite specific skills to gather data, train models and deploy them in production.

  12. Yes, but it’s not the job of a PM. A PMs job here is the same – get the best product to customers. (coming from a PM who also happens to have a Kaggle competition medal)

  13. One could argue building the right product mostly consists of setting team goals and boundaries. But I understand other organisations might not include PMs in team structure decisions.

  14. Sorry I lost the link between PM of MLstuff and structure decisions. If you are talking about capital structure decisions I would argue that PM should have an ownership of it.

  15. Agree with the rest of the commenters about factor loadings on a second order factor. Maybe this could be overcome by bifactor models but I see no point in utilizing them – just try to specify your model better, maybe run an EFA just to peek on loadings, but to measure something through that model you should establish that it can describe measurement inputs which could be done through CFA.

  16. My experience has been that there is normally a unicorn or 2 applying to most open PM jobs. These are the PMs that will get the job if they want it and no other candidate comes close.

  17. You forgot to add that “unicorness” is measured and determined subjectively.

  18. Without trust your influence attempts will only make it worse, and turn mistrust into irritation and opposition.

  19. Hey I'm looking for courses around MVP and user testing or parts where developing the idea before launch. Can you suggest some?

  20. If I ask them what a clustering is, and they say “idk, grouping and/or separating cases based on some inductive principle” – I would consider that they know their stuff

  21. Unless those items in different tests are linked you cannot really compare their parameters.

  22. Small remark: you’re describing IQ tests in the Classical test theory paradigm “add stuff up”, while specialists certainly may try to utilize this “weak learner” idea, current measures could prove their quality much better.

  23. More like when the junior start saying something to managers while he does not understand the topic well

  24. What do they know? What do I want to tell them? What would change after? Dataviz should not be created without thoughts about audience

  25. The problem is ML papers producing 100 SOTAs each day and unexperienced people trying to implement them

Leave a Reply

Your email address will not be published. Required fields are marked *

News Reporter