Jase
Reputation
610
Top tag
Next privilege 750 Rep.
See vote count; VIP usercard
 Apr6 comment How to compare different volatility measures? But what is your RV estimate? The RV function that is closest to some $VOL_i$ will make the slope coefficient larger for that particular $i$... Feb8 comment How can I go about applying machine learning algorithms to stock markets? This is thoroughly incomplete as this information may already be in the current price. You need another layer to this system to test whether the information is or is not in the price. Dec11 comment interpreting huge jumps I can't answer your question unless you give me your model. Nov9 comment Econometric vs ANN models for forecast? Brilliant answer! Oct23 comment evaluation of volatility models using loss functions Possible duplicate of quant.stackexchange.com/questions/8056/… Aug18 comment How to compute a sector's volatility within a portfolio? This question is not clear to me. If you want to calculate the volatility of the sector portfolio why is the correlation of the portfolio's constituents with other sector's constituents relevant? Aug18 comment Is variable binning a good thing to do? @xiaodai Why not? Your answer doesn't actually answer this question. Aug16 comment Is variable binning a good thing to do? This doesn't actually explain why it's better to bin. Aug15 comment Is variable binning a good thing to do? Is there a reference to back these assertions? Aug14 comment Is there a copula that can estimate negative tail dependence? Yes, you have my question right, and very nice addition! Could you please define in more detail what $(k+1)$ is? Also, is there any research on using state space modelling to get a time-varying estimate $\hat{\beta}(t)$? Aug10 comment What is the difference between volatility and variance? Variance is also associated with the underlying process (population variance vs. sample variance). Aug10 comment How do I estimate the joint probability of stock B moving, if stock A moves? The main deficiency of this would be if you want a time-varying estimator of conditional probability, since once you've got large $N$, $\hat{p}_N$ would respond very slowly to a sudden clustering of joint positive returns. Here you would want time-varying copulas (not estimated through MLE). Aug10 comment R: Fast and efficient way of running a multivariate regression across a (really) large panel (First pass of Fama MacBeth) For the non-NA matching try data[complete.cases(data),] Aug9 comment R: Fast and efficient way of running a multivariate regression across a (really) large panel (First pass of Fama MacBeth) Two ideas; (i) don't run lm(...), use $(X'X)^{-1}X'Y$. (ii) every so often do a write.csv' or a save, and rm() to clear memory, (iii) run the as.character on the whole vector of dates instead of on a single date in each loop iteration.. Aug7 comment Why are regressors squared and not ^1.5 or ^2.2 or ^2.5? @ApprenticeQueue Agreed. The parisomny argument is only coherent if the parameter value is selected when minimising the loss function. If it is set a priori then the parsimony argument doesn't hold since 1.95 is as arbitrary as 2.00. Aug5 comment Time-varying correlation via state-space representation and Kalman filter Stochastic copulas such as SCAR are complex. First, look at normal time-varying copulas in the 2006 paper by Andrew Patton. Code and paper references for the time-varying copulas can be found on public.econ.duke.edu/~ap172/code.html. Aug3 comment Rolling window Kendall's tau against APARCH(1,1) correlation Get a time-varying Kendall's tau from Patton's (2006) time-varying copula parameterisation. Heteroscedasticity in the univariate series can be modelled by specifying the marginal distributions as a GARCH process. Fat-tailed bivariate distributions can be handled using a time-varying SJC copula or alternative. Aug3 comment Time-varying correlation via state-space representation and Kalman filter You probably already know this, but as a side note consider time-varying correlations through (i) (AG-)DCC-GARCH, (ii) time-varying copulas and stochastic copulas, (iii) wavelet coherence. If you're using Kalman Filters because you want an "online" time-varying estimate of correlation, then (i) and (ii) aren't useful but (iii) with causal wavelets might be. If you want to just know what happened in the past, (ii) would be better because you can look at non-linear time-varying correlations such as a time-varying Kendall's Tau or Spearman's rho. Jul22 comment Time-series similarity measures Also, cointelation. Jul22 comment How to look for fractals/harmonics patterns in time series? Perhaps to detect fractal behaviour you could fit something like a Daubechies wavelet. That is, $W(a,b) := \frac{1}{\sqrt{|a|}} \int_{-\infty}^{\infty} f(t) \phi((t-b)/a)dt$. Then you want to check the set $\{W(a,b) : a \in \mathbb{R}\}$ where $a$ is the scale, for some fixed $b$. If all the coefficients are similar then this might be some indication of fractal like behaviour at least in that time localisation. I'm sure there's ways from stochastic calculus. Also, might be good to explore econnometric estimators (Heterogeneous ARCH) that take in multiple freq and check coefficient stability.