I thought this was pretty good. I don't have a strong background in linear algebra or statistics, but this seemed to lay out the ideas necessary for PCA in a clear manner. Looking forward to the next video.
Hey, thanks for the topic suggestion! I'm not sure yet if we will produce a video on MCA as well, but I'll keep it in mind for the future content planning. Also, thanks for the feedback regarding ads on the website. Please note that ads are currently the only way how we earn money at Statistics Globe. Creating these tutorials is a lot of work, so I think it's fair to show some ads within the content. However, we have planned to create paid courses in the future as well, so maybe this will be a better fit for you. I'll keep you updated. Regards, Joachim
Hello, I recommend you to visit our video How to Use PCA in R : ruclips.net/video/mNpBrHwOCt4/видео.html, where the implementation of PCA in R is explained in detail. Besides, you can find the whole script in the explanation. Best, Cansu
Hello Rhino, Unfortunately, I can't provide a full implementation in R for this task as it is quite complex and out of the scope of the PCA topic. I would suggest looking for research papers or textbooks on kernel density estimation that provide a detailed algorithm for these methods. It might also be possible that specific packages in R provide these methods, but I am not aware of them now. But let me give you a brief guide on how to approach this task based on my research. To compute ISE: Estimate the density using a kernel density estimation technique. Compute the difference between your estimated density and the true density at each point in your sample or on a grid of points covering the support of your distribution. Square these differences and integrate them over the support of your distribution. This can be done using numerical integration techniques. For UCV: For each bandwidth in a sequence of possible bandwidths, compute the unbiased cross-validation score. The unbiased cross-validation score for a given bandwidth is computed as follows: Compute the leave-one-out kernel density estimate for each observation in your sample. Subtract the true density at each observation from the corresponding leave-one-out estimate. Square these differences and average over all observations. Choose the bandwidth that minimizes the unbiased cross-validation score. You are also welcome to share your question in our Facebook discussion group: facebook.com/groups/737670030245660 for any further help. Regards, Cansu
The best explanation of PCA I have seen so far!! Thank you Cansu.
Hello Denny,
I am happy to hear such nice feedback. You are welcome!
Best,
Cansu
var-i-able! I cannot say this any other way now.
Haha, glad you like how we pronounce it! :D
I thought this was pretty good. I don't have a strong background in linear algebra or statistics, but this seemed to lay out the ideas necessary for PCA in a clear manner. Looking forward to the next video.
Hey Scott,
I am glad that I could catch the balance between theory and practice. Thanks for the feedback!
Best,
Cansu
Also thanks to Joachim for the useful channel.
Thank you so much for the kind comments Denny, it's great to hear that you like the video!
Thank you Joachim and Cansu for the really insightful tutorial of PCA.Looking forward to the next tutorial on this topic :)
Hey Darryl,
I am glad that you found the tutorial informative. See you in the next video!
Best,
Cansu
Thanks for the clear explanation and this helpful channel Cansu & Joachim!
Hey Yesim!
You are very welcome! I hope you also like the upcoming ones :)
Best,
Cansu
Thanks a lot.
You are very welcome!
Best,
Cansu
Thank you! Will you also do a series on independent component analysis? Thanks in advance
Hello Griffith,
You are welcome! For now, we don't have such a series. But we will consider your feedback for our future tutorials.
Best,
Cansu
Great timing! Any chance you can do one on MCA also? Also, your website had progressively gotten worse with ads that it is borderline unusable
Hey, thanks for the topic suggestion! I'm not sure yet if we will produce a video on MCA as well, but I'll keep it in mind for the future content planning. Also, thanks for the feedback regarding ads on the website. Please note that ads are currently the only way how we earn money at Statistics Globe. Creating these tutorials is a lot of work, so I think it's fair to show some ads within the content. However, we have planned to create paid courses in the future as well, so maybe this will be a better fit for you. I'll keep you updated. Regards, Joachim
I like this new type of videos, keep up the good work please 😊
Hello,
It is great to hear! Thank you for your encouraging words.
Best,
Cansu
Help with interpretation of biplot
Hey, please have a look here: statisticsglobe.com/biplot-pca-explained
Thank you for excellent explanation of PCA. Can you please share the R script for this lesson?
Hello,
I recommend you to visit our video How to Use PCA in R : ruclips.net/video/mNpBrHwOCt4/видео.html, where the implementation of PCA in R is explained in detail. Besides, you can find the whole script in the explanation.
Best,
Cansu
I need help!!!
How can i program the integrated squared error (ISE) of Gamma distribution using the Gamma kernel with (UCV(h)) in R???? PLEASE 🙏
Hello Rhino,
Unfortunately, I can't provide a full implementation in R for this task as it is quite complex and out of the scope of the PCA topic. I would suggest looking for research papers or textbooks on kernel density estimation that provide a detailed algorithm for these methods. It might also be possible that specific packages in R provide these methods, but I am not aware of them now.
But let me give you a brief guide on how to approach this task based on my research.
To compute ISE:
Estimate the density using a kernel density estimation technique.
Compute the difference between your estimated density and the true density at each point in your sample or on a grid of points covering the support of your distribution.
Square these differences and integrate them over the support of your distribution. This can be done using numerical integration techniques.
For UCV:
For each bandwidth in a sequence of possible bandwidths, compute the unbiased cross-validation score.
The unbiased cross-validation score for a given bandwidth is computed as follows:
Compute the leave-one-out kernel density estimate for each observation in your sample.
Subtract the true density at each observation from the corresponding leave-one-out estimate.
Square these differences and average over all observations.
Choose the bandwidth that minimizes the unbiased cross-validation score.
You are also welcome to share your question in our Facebook discussion group: facebook.com/groups/737670030245660 for any further help.
Regards,
Cansu