2,649 reputation
514
bio website
location
age
visits member for 2 years, 5 months
seen 11 mins ago

Oct
16
comment Expected Shortfall (CVaR) Backtesting
@emcor I had meant simply using the historical CVaR for the purposes of the backtest, rather than rely on any assumptions about the distribution of returns for the strategy. If you want a confidence interval on the ES, you can bootstrap from the historical returns. I've seen some literature on expectiles, but I haven't had the chance to read it yet.
Oct
3
comment Portfolio Turnover Constraint
@Richard It's not just them being positive, it's the piecewise nature of the constraint. Optimizers that require continuous functions will tend to not like it when you use absolute values. Also, this approach can easily be extended to include transaction costs, though you probably need to add in non-linear constraints $b_{i}s_{i}=0$ to ensure that sells are zero if you have buys, and vice-versa. My recollection is that it's not usually needed for turnover, but there might be cases where it is.
Sep
30
comment How to interpret ACF and PACF plots
I don't see anything interesting in these. Looks stationary.
Sep
25
comment Factor model assumptions
Alternately, consider a model like Barra where you use cross-sectional regression. The $B$ term changes in every period, but you can still use that formula to decompose the covariance of $X$.
Sep
25
comment Factor model assumptions
Suppose you regress every stock in $X$ against the Fama-French factors (so this is a time series factor model). If you assume constant 2nd moment, then you can do the covariance matrix of $X$ as is normal for factor models. If you assume Garch volatility for $F$ with constant correlation and assumptions as above for idiosyncratic returns, then you can still use that formula, but you have to adjust it to be a conditional covariance.
Sep
15
comment Exporting Time Series Data For Securities Prices From Bloomberg to Excel
Just get your firm to pay for the add-in.
Sep
9
comment How do Return.portfolio and Return.rebalancing work in Performance Analytics in R?
I do not see your motivating example in the PerformanceAnalytics pdf. The closest comparable thing I see is round(Return.rebalancing(edhec,weights),4). You'll note that the 2007-01-01 return matches up with edhec[121,1:11]%*%t(weights[8,]), ignoring the rounding.
Aug
28
comment What kind of front end/ gui is used with trading applications?
Can you recommend a more thorough reference on the subject (preferably starting from the basics and working up)?
Aug
28
comment State Space models with Short Time Series
More to just get a sense if it's working the way that it should.
Aug
28
comment State Space models with Short Time Series
I was focusing on the estimation. Combining that with testing a trading strategy will introduce more complexity. No one right answer, I think. One thing I do sometimes is assume the model is true, then simulate a bunch of paths of data, and then run the trading strategy on simulated data and evaluate it way on the larger sample. Not sure if that makes sense in your case, because I'm really not sure what you're doing.
Aug
28
comment Fixed volatility portfolio with max returns creates skewed results
Yes, it is possible. Assuming you did the non-linear constraint correctly, it could be that with four assets it can't find an optimal portfolio with volatility that low. There are too many potential things it could be to say for sure. Some other tips, draw the efficient frontier (that way you can get a sense of the minimum risk and maximium return portfolios) or switch it around and minimize risk given return (and maybe also draw that frontier).
Aug
27
comment Fixed volatility portfolio with max returns creates skewed results
It is likely an error that can only be evaluated by looking at the code. For instance, it could be that you programmed the non-linear constraint wrong, or it could be that your linear and non-linear constraints are inconsistent somehow. One piece of advice: start with a very simple optimization and gradually make it more complicated, verifying at each stage that you're getting sensible results.
Aug
26
comment State Space models with Short Time Series
Without having a better sense of what you're trying to do, things I would try are 1) increasing frequency of data, 2) leave one out cross validation, 3) Bayesian methods.
Aug
18
comment factor models and using cross section regression
@Viquar I'm more familiar with the approach in that paper as applied to momentum than I am with respect to characteristics-based factors (e.g. P/E). In a cross-sectional approach (as in Barra), one can use a GLS approach to account for the correlation between factors, though I'm not sure how important it is for the level of coefficients versus standard errors.
Aug
12
comment Why maximize expected growth rate?
@RRL Wrt to your first paragraph, it might make it a bit more clear to say that Kelly is maximizing the log of wealth, which is equivalent to saying utility is the log of wealth.
Aug
6
comment Johansen Cointegration Test
@user3126171 So perhaps your problem is that you don't understand cointegration or the Johansen test, rather than necessarily how it is implemented in Matlab. As I said, the Spatial Econometrics toolbox manual is very clear about what the Johnansen test means spatial-econometrics.com/html/mbook.pdf Otherwise, most books on time series econometrics tend to cover it. You can also just google johansen test and there are many results that explain it.
Aug
6
comment Johansen Cointegration Test
Honestly, I tend to use the one in the Spatial Econometrics toolbox (its manual is a bit clearer, though there's a link at the bottom of the Matlab link I have above that provides more details). The Matlab test is testing r=0, r=1, i.e. whether the number of cointegrating relationships is some number. The test rejects r=0, so you can say there is one cointegrating relationship.
Jul
31
comment Why are we obsessed over normalizing financial data?
@BCLC I would consider standardizing a type of normalizing, but most of the time when people talk about normalizing they're talking about standardizing.
Jul
31
comment Why are we obsessed over normalizing financial data?
@BCLC Well you could use logs or log returns with economic and financial data. Depends on what you're doing.
Jul
29
comment Why are we obsessed over normalizing financial data?
Normalization typically means subtract the mean and divide by the standard deviation. That transformation won't make non-normal data normal.