Comments on: About

By: Anonymous

Anonymous — Fri, 15 Dec 2023 19:10:52 +0000

Hi Sooper useful. Especially the PCA analysis explanation. Thx very much

By: ejc

ejc — Mon, 20 May 2019 16:26:56 +0000

re: your write up on PCA, you suggest using PC3 as the hedge ratio, i’ve read elsewhere that people try to hedge out PC1 and PC2, s.t. their hedge ratios are the product of PC1*PC2. It turns out that PC1*PC2 ~ PC3 – i was wondering if you had any insight or thoughts on that relationship and the two approaches.

I was thinking that since Vector multiplication results in a Vector that is Orthogonal to both and the idea behind PCA is to calculate orthogonal vectors in descending significance that it makes sense that PC3 would be ~ PC1*PC2.

LikeLike

By: anubis (@upperouestside)

anubis (@upperouestside) — Sat, 16 Mar 2019 15:40:46 +0000

Recently discovered this treasure trove of a resource. Firstly, my genuine and sincere thanks for sharing these incredibly useful insights and explorations.

With the vantage point of your experience and hindsight, I’m curious to hear your thoughts as to how successful (PnL-wise) and applicable these approaches are when married with a discretionary macro mandate (as opposed to purely systematic). I suspect an unspoken postulate that runs through these posts is that they don’t really ‘work’ (again, PnL-wise) on a ‘purely’ systematic basis (please do let me know if you disagree.) I’m curious as to how complementary, applicable and useful you found these methods when used as part of a quantitatively informed discretionary heuristic.

Also curious as to whether you’ve dabbled in Python, given a lot of these methods port quite nicely with it.

Would be more than happy to take this discussion ‘offline’ via DM, email or twitter.

Best –

LikeLike

By: Annie

Annie — Tue, 30 Jan 2018 22:23:30 +0000

Hi bQuant,
Thanks for a great series on SVM. I have one question. In the excel workbook svm_kernel_calcs_example, tab = Implicit Feature Mapping (Kernel) the solver determines the Alphas (H6:H34) and the value for coefficient b (R5). Q. how would you use these solved values to predict a new value, say X1 = 0 and X2 = 0.5 ?
Annie

LikeLike

By: David Wood

David Wood — Wed, 17 Jan 2018 18:17:20 +0000

I have spent some time working through your three part tutorial on SVM, and I have learned a lot. Thank you. However, in trying to apply it, i need to extend it to multiple input dimensions, not just xi and x2 but x1, x2, x3, x4 etc, Although I think I now understand how to get the alphas for such cases, I cannot work out how to define a formula for the multi-dimensional hyperplane, or place a new test case into a formula to obtain its x1 value knowing x2,x3,x4 and alpha etc. I think it would be great if you could post a part four to your SVM tutorials addressing multi-dimensional input case. Or, alternatively, simple explain what I am failing to grasp about multi-dimensional cases.

LikeLike

By: Michel

Michel — Fri, 05 Jan 2018 20:41:02 +0000

Hi, Am trying the codes on logistic regression with my set of data but am getting ‘#value’ error. what could be the problem. Any help please

LikeLike

By: ab crombie

ab crombie — Thu, 17 Nov 2016 03:21:24 +0000

In reply to bquanttrading.

hey bqquant, thanks for the tips, I can now say I am a huge fan of R, partially thanks to you. even if I cant understand how you wanna do a lot of the math stuff still in Excel.

I just had a quick question i was hoping you might have had experience with or have so simple ideas. I am looking to make a score of various economic stats, like Housing, Retail Sales, Employment. large categories with a few different stats to consider. I would like to make a simple score say on a scale of 1 -10 or -10, +10. Any thoughts, ideas or research you may have seen.

I was thinking of doing a composite some score based on the distance of some moving average and a score referring to the short term trend./ monthly change

LikeLike

By: bquanttrading

bquanttrading — Fri, 22 Jul 2016 03:39:16 +0000

In reply to ab crombie.

hi abc, I thought A Ng’s coursera MOOC was great. Its probably important to refresh on the math. Just grab a couple used textbooks and use them as ref when plowing through the course. Intro to stat learning https://lagunita.stanford.edu/courses/HumanitiesSciences/StatLearning/Winter2016/about is very good. I also liked Carol Alexander’s Market Risk Analysis series http://ca.wiley.com/WileyCDA/WileyTitle/productCd-0470998008.html. If you want to learn excel/vba the best way i think is to just pick a project and work through it and use the internet. I would strongly recommend http://ca.wiley.com/WileyCDA/WileyTitle/productCd-0471387347.html and http://shop.oreilly.com/product/9780596008796.do. Theres enough here for you to see what you find interesting and what your strengths are and continue from there.

gluck

LikeLike

By: ab crombie

ab crombie — Thu, 07 Jul 2016 18:27:47 +0000

great blog! I really appreciate all the effort and complete posts. Quick question, for someone with a CFA but not an advanced mathematics or CS degree, what are your thoughts on the best way to get up to speed in the quant world. Any coursera courses/books strike you as really good. I did Ng’s Machine Learning but couldnt keep up with both the math and programming. any thoughts would be appreciated. I see your platform of choice is excel and R

LikeLike

By: bquanttrading

bquanttrading — Sun, 29 May 2016 09:43:53 +0000

In reply to Kimberly Lanato.

i think the best way to learn is to try it yourself. besides, its likely you will find a bug here and there if you try to replicate the work on your own. also, there is no way to upload macro enabled files on wordpress. there may be a hack but i couldnt be bothered to look it up.

LikeLike