401 reputation
28
bio website
location Oslo, Norway
age
visits member for 3 years, 1 month
seen Nov 12 at 8:06

Dec
9
comment Calculating log returns using R
There was a slight (obvious?) error in my reply with taking the negative logreturn instead. Now corrected. Thanks to user6569 that notified me.
Jul
12
comment how to calculate more efficient volatility figure than historical volatility?
Does your question imply that you do not have access to option market data you can derive implied volatilities from? (i.e. you only have the underlying market?)
Nov
17
comment Why is random trading minus transaction costs not zero expected value?
Would you consider the bid ask spread be considered a transaction cost? If not, a very simple counterexample could be provided. But I assume that is not what you are asking about?
Sep
12
comment Correlation: Test for linear dependence
Thanks. I think the second paper maybe can solve some of my issues (restated in a comment above). I will have a look at it.
Sep
12
comment Correlation: Test for linear dependence
I have started to be a bit vary of just looking at the correlation coefficients, as they are simply a number and could be stable although the returns do not follow a lognormal (or any other assumed distribution) walk with correlation p. I guess one would have to look at the distribution as a whole, considering whether the observed data would be sampled from e.g. a bivariat normal distribution.
May
11
comment age-sensitive correlation measurements in finances
Have a look at RiskMetric's "Technical Document" from 1994, it should all be explained there. It also contains a recursive formula :) If not, the r's in the formula is logreturn of stock j at time (t-n), while your lambda is the weighting constant (RiskMetrics used 0.94, and this usually works well)
Apr
24
comment How to get greeks using Monte-Carlo for arbitrary option?
Numerical derivatives are iffy business, but I agree that it seems to be your best choice. As you probably know; be aware of the how the precision decreases quickly(!) as higher orders are measures.
Apr
20
comment Analytical relationship between a covariance matrix and cross-sectional dispersion
And when you measure standard deviation, are you using the estimator 1/(N-1) * sum(r_i,t * r_i,j) (summed over some time)
Apr
20
comment Analytical relationship between a covariance matrix and cross-sectional dispersion
so if r_i and r_j are the returns of each stock, you are looking for the expected value of the product of these two? i.e. E(r_i*r_j) ?
Apr
20
comment Do I need a copula to accurately estimate the VaR of a portfolio of risky assets?
@Alexey Kalmykov Of course, but if you choose method and distribution right, I dont really see how a historical approach will be better than a distributional.
Apr
20
comment Do I need a copula to accurately estimate the VaR of a portfolio of risky assets?
Historical VaR will not measure events that "have not already happened" in your data set. Hence, you will get a more general result if you do some distributional assumptions.
Mar
5
comment Quantitative Analysis Games on Investing?
Does this mentioned game above come in a non-finish version? The documentation looks fun, but when I try to navigate around I get to finish pages.
Feb
20
comment How to minimize the difference between a parametric VaR and a MC-VaR with lognormal assumption?
As I have not found any good litterature on linear combination of lognormals and this question round neither gave any good results, I assume it is because there is no currently known closed solution of the problem stated. A hope for the question was to find something like a "better approximation" to the problem. I guess the best solution is to either reduce number of simulations or accept the (unknown) deviation. Thanks for help, guys.
Feb
8
comment How to minimize the difference between a parametric VaR and a MC-VaR with lognormal assumption?
Setting a fixed seed will reduce the "mc-noise" but the two methods will still differ as we simulate prices as lognormal, but use that the arithmetic return is normal in the parametric case. Your second approach is smart but i have already hiven it a try. As the portfolio changes over time too, it didnt come out very well in my "sample tests"...
Feb
8
comment How to minimize the difference between a parametric VaR and a MC-VaR with lognormal assumption?
True, but hen you have the multivariate case. Do you know how linear combination of lognormal distributed rv's looks like? The litterature Ive seen looks messy. Unfortunately..
Feb
8
comment How to minimize the difference between a parametric VaR and a MC-VaR with lognormal assumption?
Yep, think the method goes under that name too. The problem is not to get a better model in that way, rather it is to find a closed form solution that mimic the geometric part of the stock model.
Feb
7
comment What weights should be used when adjusting a correlation matrix to be positive definite?
Well. Thinking within the model, one would believe that some correlation pairs are more trustworthy than other (e.g. high liquidity implies good data implies a better correlation coefficient). Thinking outside a model, looking at how much the correlation have changed in history could also be a parameter that determine the weight of a pair (high variability in time implies lower weight). Not sure if its worth an own question. Rather it is a part of the overall, although the quesiton above consist of technical and a economical considerations.
Feb
7
comment What weights should be used when adjusting a correlation matrix to be positive definite?
Yes, I agree, but I have received some pretty nasty results while weighting. Also, how should one evaluate correlation pairs?
Feb
6
comment Calculating log returns using R
hadnt noticed the "diff" function yet. A handy one, indeed.
Nov
8
comment What weights should be used when adjusting a correlation matrix to be positive definite?
Thanks. Well, so far I have not found any solution and are currently running unweighted approximations. I find this about alright, but as I am approximating correlations from some stocks that are somewhat illiquid it would be satisfying knowing that these will be altered more than the main stocks in our portfolios..