If someone just using equation and abstract dataset, I will be completely lost, so I really like the way you explained the concept, by using detailed and realistic dataset, with such plenty and friendly arrows to indicate the positions we should focus, it significantly help me to get fully understanding of it.
Man, that's a great explanation. I'm preparing for a test at the uni regarding logostic regression and you saved me a lot of time I would need to spend on reading a statistics book.
Hi, thank you so much for the great videos. Quick question about the interpretation of residual deviance. I understand the lower the value, the better. Is it because the LL (proposed model) inside the residual deviance formula follows the same principle? Thanks in advance
Yes, the lower the LL is, the better the model fits the data. Maybe my latest video about MLE vs OLS is of interest... ruclips.net/video/bhTIpGtWtzQ/видео.html
@@tilestats I didn’t get your answer notified. Sorry I really didn’t mean to be late in thanking you. One other doubt I have is about plotting a multi variable logistic regression model with the predicted probabilities on the y axes. I’m trying this in R, but I’m a bit confused on how to go about this
Great job! So, the null model of logistic regression is essentially always that both outcomes have an equal probability of happening? The null model is it's all due to chance?
No, the null model, in this case, is the sample proportion of "successes". Since I had an equal number of cancer patients and healthy individuals, in this example, the null model is equal to a horizontal line with an intercept of 0.5. However, if we would have a larger proportion of cancer patients, the null model will have an intercept greater than 0.5. Also, the null model may sometimes represent the null hypothesis, which can be a model with several parameters (see next video).
Did you watch the section around 10:00? If so, to get the parameter values that results in the lowest negative log-likelihood, you can for example use Gradient descent: ruclips.net/video/31w-xQX0Z_8/видео.html Or simply use a statistical software tool that will directly give you the estimated values.
You have done a fantastic job explaining the details! Thank you! Your video is one of the top videos on RUclips on the topic.
Thank you!
If someone just using equation and abstract dataset, I will be completely lost, so I really like the way you explained the concept, by using detailed and realistic dataset, with such plenty and friendly arrows to indicate the positions we should focus, it significantly help me to get fully understanding of it.
Man, that's a great explanation. I'm preparing for a test at the uni regarding logostic regression and you saved me a lot of time I would need to spend on reading a statistics book.
Thank you and good luck with the analysis.
Very good and clear explanation, 1:28 The P(Cancer) nominator should be e(BX)
Very very good complement material for an analysis course I am taking! Thank you soooo much!
Thank you!
Huge thanks
Can we say the pseudo-r-square is the same as deviance ratio that is reported in some statistical packages after logit models
best explaination
Thank you!
Hi Sir,
Please provide videos gradient descent and ascent in case of possible.
Which can help us more.
Hi, thank you so much for the great videos. Quick question about the interpretation of residual deviance. I understand the lower the value, the better. Is it because the LL (proposed model) inside the residual deviance formula follows the same principle? Thanks in advance
Yes, the lower the LL is, the better the model fits the data. Maybe my latest video about MLE vs OLS is of interest...
ruclips.net/video/bhTIpGtWtzQ/видео.html
@@tilestats I didn’t get your answer notified. Sorry I really didn’t mean to be late in thanking you. One other doubt I have is about plotting a multi variable logistic regression model with the predicted probabilities on the y axes. I’m trying this in R, but I’m a bit confused on how to go about this
thanks for the great video! at 11:32 why b0 is 0 at null model? could you please illustrate? many thanks!
Great job! So, the null model of logistic regression is essentially always that both outcomes have an equal probability of happening? The null model is it's all due to chance?
No, the null model, in this case, is the sample proportion of "successes". Since I had an equal number of cancer patients and healthy individuals, in this example, the null model is equal to a horizontal line with an intercept of 0.5. However, if we would have a larger proportion of cancer patients, the null model will have an intercept greater than 0.5. Also, the null model may sometimes represent the null hypothesis, which can be a model with several parameters (see next video).
@@tilestats Thank you!
I can't understand how obtain -5.754 and 2.747.
Did you watch the section around 10:00? If so, to get the parameter values that results in the lowest negative log-likelihood, you can for example use Gradient descent:
ruclips.net/video/31w-xQX0Z_8/видео.html
Or simply use a statistical software tool that will directly give you the estimated values.
Uhmm sir , my proff. Stole your slides and he is teaching us of them is that legal?
As long has he/she acknowledge the source, it is ok.
I still don't know how b0=-5.747 and b1=2.747. I used cancer=1 and healthy=1 on the dataset and i have b0=1.5 and b1=1.25
Healthy should be zero. What software do you use?
@@tilestats i use casio calculator
and yes i use healthy =0, i miss type
Can you do logistic reg on a calculator? Imstall R instead so that I can give you the code.
Your videos are really help my understanding, but I still got the wrong value. Can you please share the r code?