Wow, thanks Martyr022! I appreciate the feedback. I can definitely work on making more.. I've been playing around with using the white board you see with this video... though its kinda hard to see my small handwriting? My normal style video with a black background and drawing computer app would probably be better... what do you think?
@@Stats4Everyone Personally I do prefer those other videos with the black background and drawing via the computer app, but I also had no issue following along here as well. Looking forward to more vids!
@@Martyr022 awesome. Me too thanks! thanks for the feedback! If there is a particular example, or proof, you'd like to see, please always feel encouraged to share :)
@@Stats4Everyone I know I said I look forward to more Logistic Regression videos (and I still do) but in general I would be interested in videos on the Generalized Linear Models for data such as counts, ordinal data (so probit/logit models) and the like One thing that I just always have a hard time wrapping my head around are Moment Generating Functions--I don't know why but there's something about it that just doesn't click with me. So some videos on that would be cool
Just came here after seeing your another video on the similar topic. It would be great if you could make a video on the regularization of logistic regression. Thanks :)
This is a good question Taotao! My intuition would be that yes, you can get caught at local maximum, and the algorithm may miss a global maximum. To look into this question further, we should find the second derivative of the log likelihood function. Is there an inflection point (where second derivate is zero), or is the log likelihood strictly concave (second derivative is less than zero)? I would need to look into this further. It is a good question, and my initiation is that it might depend on the data, though again, I would solve the second derivative to explore this question more.
@@Stats4Everyone also it’s actually very nice that the 2nd derivative could be used as fisher information matrix for quantify uncertainty of the estimator. Would it be possible that you can make a video to explain this? I am sure a lot of people will be interested in this topic.
![Megan Thee Stallion - Bigger In Texas [Official Video]](http://i.ytimg.com/vi/QxTkZTLWdo4/mqdefault.jpg)
REALLY good video! I hope to see more videos on logistic regression in the future
Wow, thanks Martyr022! I appreciate the feedback. I can definitely work on making more.. I've been playing around with using the white board you see with this video... though its kinda hard to see my small handwriting? My normal style video with a black background and drawing computer app would probably be better... what do you think?
@@Stats4Everyone Personally I do prefer those other videos with the black background and drawing via the computer app, but I also had no issue following along here as well.
Looking forward to more vids!
@@Martyr022 awesome. Me too thanks! thanks for the feedback! If there is a particular example, or proof, you'd like to see, please always feel encouraged to share :)
@@Stats4Everyone I know I said I look forward to more Logistic Regression videos (and I still do) but in general I would be interested in videos on the Generalized Linear Models for data such as counts, ordinal data (so probit/logit models) and the like
One thing that I just always have a hard time wrapping my head around are Moment Generating Functions--I don't know why but there's something about it that just doesn't click with me. So some videos on that would be cool
@@Martyr022 Great suggestion! I can definitely make some videos about Moment Generating Functions. I wrote it down on my to do list :-)
Just came here after seeing your another video on the similar topic. It would be great if you could make a video on the regularization of logistic regression. Thanks :)
Really awesome! Is this score a convex function (local maximum possible or no) ?
This is a good question Taotao! My intuition would be that yes, you can get caught at local maximum, and the algorithm may miss a global maximum. To look into this question further, we should find the second derivative of the log likelihood function. Is there an inflection point (where second derivate is zero), or is the log likelihood strictly concave (second derivative is less than zero)? I would need to look into this further. It is a good question, and my initiation is that it might depend on the data, though again, I would solve the second derivative to explore this question more.
@@Stats4Everyone awesome and thanks for explaining! Would be very appreciated to have you explain this if you figure it out.
@@Stats4Everyone also it’s actually very nice that the 2nd derivative could be used as fisher information matrix for quantify uncertainty of the estimator. Would it be possible that you can make a video to explain this? I am sure a lot of people will be interested in this topic.
@@taotaotan5671 Yes, this does sound like a fun idea. I'll respond here again once I have a video. Maybe later next week. Thanks!
Really good ! Cheers