2,296 reputation
312
bio website
location
age
visits member for 1 year, 11 months
seen 8 hours ago

Jan
17
comment Inflation modelling
With respect to Richard's point, the European HICP inflation data all comes out NSA (non-seasonally adjusted). I checked and the Spanish data from INE is also NSA. You just have to be careful because other countries might provide the SA data (like the US does). Spanish prices are typically weak in winter before recovering in the spring.
Jan
9
comment How can I calculate the Cumulant-Generating Function in Matlab?
Meucci might have code that does this. mathworks.com/matlabcentral/fileexchange/authors/21105
Jan
6
comment Optimizing Principal Component factor weightings over time
If you have a time-varying covariance matrix (you could construct with Garch volatility and either a constant or time-varying dependence structure), then you can perform PCA on each period or project out to the future. Not sure if that's what you're looking to do.
Jan
6
comment Expected Shortfall (CVaR) Backtesting
Why don't you just try a backtest and to calculate ES and see if it works? I use CVaR all the time in backtests, though I mostly do it with monte carlo simulations.
Jan
6
comment Lagged dependent variable, yes or no?
I probably would probably first emphasize the calculation of standard errors in the presence of autocorrelation before anything like omitted variable bias.
Dec
20
comment How to group timeseries showing similar curve
Do you ever throw in regression coefficients, like regression against indices (including sector/industries)?
Dec
12
comment Optimization: Factor model versus asset-by-asset model
However, if the weight constraints are too restrictive, then it might not work as you'd like.
Dec
12
comment Optimization: Factor model versus asset-by-asset model
I'm not so sure that I've read that portfolio optimization on $\widehat{\Sigma}$ is "easier" than the optimization on $\Sigma$. You're still working with an $N\times N$ matrix, but you've ensured that it is positive definite (which is a good thing). You could always generate some random data and test the idea. It might be possible to set up the optimization in terms of principal portfolios so you're only operating on the first few $K$ principal portfolios and then translating those weights back into normal weights for constraints and TC.
Dec
12
comment What is Quantitative Investing and how does it differ from Quantitative Trading?
Most of the quant AUM is long-only (even if they build long-short factor models internally). In my view the quant managers have better risk control than many other active managers. The good ones can achieve comparable alpha with lower tracking error, giving them higher information ratios. Also, my understanding is that the fees are typically lower for quant managers than other active managers.
Dec
5
comment Large (5K-10K) non positive definite (particularly near singular) covariance matrices and treatments for Cholesky decomposition
If you're working with futures contracts, perhaps you should begin by reformulating the problem. For instance, you could start with a parsimonious model of the futures curve (akin to the Nelson Siegel yield curve decomposition) and do that for each period. This way you could reproduce any futures curve with just a few variables. So you'd then only construct the covariance matrix using the time series of the factors from all the futures curves.
Nov
23
comment Potential pitfalls in the use of correlation
The price indices for fixed income (obviously) don't incorporate coupon payments so any changes in price should be due to movements in yields (which is why I prefer to model the YTMs than the price or total return).
Nov
21
comment Components of an index in a specific date
I just use whatever my company subscribes to. I don't know for sure. My recollection is that Factset compiles some data on their own (such as by downloading economic data from government websites). Index composition data often comes from vendors (e.g. S&P or MSCI) and is typically quite expensive.
Nov
21
comment Components of an index in a specific date
It depends on what you subscribe to. For instance, PA2 can do attribution in Factset. Attribution requires historical index composition. I have found it easier to download index members from Bloomberg than Factset, but it may just be that my firm doesn't subscribe to whatever Factset service makes it easy to download that information.
Nov
21
comment Components of an index in a specific date
It's not just the components of the index, but you probably also would want the weights in the index. It depends on the index, but you usually have to pay for many. Most people typically get them through licenses that their company pays for (and then they would access the data through Bloomberg/Factset/ThompsonReuters/etc).
Nov
5
comment How to properly take averages to reduce data in regression/panel data analysis
@Richard The OP might get a better answer from there (especially if she makes it more clear what she's trying to do). More generally, so-called big data issues are important for quants (and if she had phrased the question about unbalanced panel methods with huge financial datasets, then that might be a good question here).
Oct
21
comment GARCH(1,1) prediction in R - Basic Questions
That method for annualizing volatility only works under a restrictive set of assumptions.
Sep
26
comment What is the correct Stutzer index and Sharpe ratio relation, assuming a normal returns distribution?
I found a better definition here: activetradermag.com/index.php/c/Article_Follow-ups/d/…
Sep
18
comment Robust Returns-Based Style Analysis
You might want to make this question a bit more specific, such as listing what type of analysis you want to perform or the types of questions you want to be able to answer (e.g. what is a robust approach to know how my exposure to x changed over time).
Sep
12
comment Testing the validity of a factor model for stock returns
You might make it a little more clear using $i$s for the cross-sectional index and $t$s for the time index.
Sep
3
comment Why non-stationary data cannot be analyzed?
I think the (almost) always difference advice can be misleading. It's important to know why and when they should difference or not. For instance, if they were to difference interest rates, then you would be ignoring the longer-run mean-reversion. Estimating an AR model in differences wouldn't help matters. From the perspective of an AR model, it is only the dependent variables that would need to be differenced for the t statistics to make sense. Lags could still be in levels, allowing for mean-reversion effects.