Reputation
3,200
Top tag
Next privilege 3,500 Rep.
Protect questions
Badges
8 16
Newest
 Yearling
Impact
~67k people reached

2d
comment Is the Altman Z-Score broken?
You are evaluating a model created to predict bankruptcies (in manufacturing firms) by testing whether it can predict returns. Does not make sense to me.
May
21
comment Is R being replaced by Python at quant desks?
On statistics, Python just doesn't have as much developed as R does on those fronts, but statsmodels is my goto. One option is to just call R functions with Python (RPy, but not apparently well tested with Windows). On derivatives, pyql seems to be more developed than the R version. On machine learning, scikit.learn. One other benefit of Python is that iPython Notebook is better than Rstudio. However, the Jupyter project looks interesting (and should work with R).
May
21
comment Is R being replaced by Python at quant desks?
I'm more-or-less with @vonjd on this. I've used a bunch of languages, but I'm most productive for statistics and analytics in R. Python is a great language and numpy and pandas are a fantastic combination. However, the community for R is just so much better. I once spent several hours over a few days to get ipopt to work on Python, and then it just worked without any real effort with R. I also don't like the conda package manager because I can't get it to work behind a corporate firewall.
May
21
comment How to identify the orders p and q for ARIMA model using least squares method?
You might find this informative: coolstatsblog.com/2013/08/14/using-aic-to-test-arima-models-2
May
21
comment R package for portfolio
@AntoniusGavin What you're really hoping is that someone is going to hide the implementation details from you. The way all these optimizers work is by implementing some algorithm(s). These algorithms work for some types of problems and not for others. An algorithm that can solve a simple quadratic problem can do it very fast and it is better just to use that. However, for your type of problem a standard quadratic optimizer will not work. So you have to use a different algorithm. nloptr allows you to choose one that works (slsqp).
May
21
comment Z-Score calculation for a win-loss streak
@Lyrk Glad you figured it out.
May
20
comment R package for portfolio
Good answer. nloptr would allow either an L1-norm or an L2-norm.
May
18
comment constrained portfolio optimization by fmincon
As a general rule, I would recommend starting with the simplest possible case and then making it more complicated. The only thing that sticks out to me is that when you use the portrisk function in fmincon you may not be passing the covmat variable with it. See the answer here: stackoverflow.com/questions/18946407/…
May
13
comment Regressing NYSE returns: Lagged intercept term & efficient market hypothesis
Oh yeah, and the significant coefficient on the first slope parameter should be negative over a long period of time, consistent with the one month reversal effect.
May
13
comment Regressing NYSE returns: Lagged intercept term & efficient market hypothesis
I agree with Quadtopic about the test. Also, I used weekly S&P 500 data to test this. In the first equation, I get that the first two parameters are significant. It's just that the quadratic one isn't. I get the same result in the second equation. I don't see what the quadratic terms get you. I'm disinclined to include quadratic terms in a regression unless I have a clear reason to do so.
May
10
comment How to deal with missing returns when creating value (equal) weighted returns
Yes. I'd be worried about delisted firms, but also firms that go bankrupt.
May
9
comment What are the main market anomalies/inefficiencies detected in quantitative finance?
Here's a recent paper detailing over 80 anomalies: papers.ssrn.com/sol3/papers.cfm?abstract_id=2508322
May
8
comment What are the main market anomalies/inefficiencies detected in quantitative finance?
@Quantopic vonjd presently has the top answer on a post discussing the difference between risk factors and anomalies. I'm not a particularly big fan of the answer, but it's consistent with academic finance. To me, it doesn't really matter what they're called. Size and value used to be considered anomalies by some, but now the profession calls them risk factors. Oh well.
Apr
18
comment Monte Carlo simulation returns not normal distributed
If you divide a lognormally distributed variable by a constant, then it will still be lognormally distributed.
Apr
17
comment Monte Carlo simulation returns not normal distributed
Geometric Brownian motion is lognormally distributed in levels. en.wikipedia.org/wiki/Geometric_Brownian_motion
Apr
15
comment Is a stationary process necessarily mean-reverting?
Seems weakly stationary to me.
Apr
9
comment Inferences with non-normal data
I understood what you meant, and I answered all of your questions. In regression, the assumptions are more related to the errors than the actual data. You can use Newey-West if you're worried about heteroskedasticity or autocorrelation. You don't need to make any adjustments for normality to make inferences, but there are techniques you can use regardless.
Apr
7
comment Inferences with non-normal data
Do you mean percent changes of index closing values?
Apr
6
comment Please give a step-by-step explanation on how to build a factor model
You're asking for links after not asking for links....All the references the writer used are to classic papers or one of the most well-known finance textbooks.
Mar
12
comment MLE estimate of normal distribution
The MLE variance estimator for normal distributions is biased because it divides by $n$ rather than $n-1$, see ee.columbia.edu/~dliang/files/mle_biased.pdf. Not sure how much that relates to the above quote.