Serving the Quantitative Finance Community

Search found 438 matches

by Rez
August 4th, 2008, 9:41 pm
Forum: Technical Forum
Topic: not i.i.d. log-returns
Replies: 8
Views: 51922

not i.i.d. log-returns

Levy processes are non Gaussian, but still IIDI guess you should start with stochastic volatility modelsK
by Rez
August 4th, 2008, 2:55 pm
Forum: Technical Forum
Topic: correlation in one factor/two factor
Replies: 1
Views: 50609

correlation in one factor/two factor

<t>in general, the bond price is a function of the short rate and maturity P(r(t),t,t+T)and r(t) is a one-dimensional diffusion which depends on dWapplying Ito's formula for two different maturities will give dP(r(t),t,t+T1) = drift1 + vol1 * dWdP(r(t),t,t+T2) = drift2 + vol2 * dWand the instantaneo...
by Rez
August 4th, 2008, 12:02 pm
Forum: Technical Forum
Topic: estimate Hull White model parameters
Replies: 23
Views: 158620

estimate Hull White model parameters

hhIn the attached document (zipped PS) there is an example of HW tree implementation in Matlab.It should be clear how the algorithm worksKyriakos
by Rez
July 31st, 2008, 8:23 am
Forum: Technical Forum
Topic: Negative Rates - Gaussian 2 factor Model (D2++)
Replies: 4
Views: 52032

Negative Rates - Gaussian 2 factor Model (D2++)

To get the ball rolling: regularizeK
by Rez
July 31st, 2008, 8:20 am
Forum: Technical Forum
Topic: ARMA(1,1) and risk neutral measure
Replies: 1
Views: 50761

ARMA(1,1) and risk neutral measure

<t>QuoteOriginally posted by: TannesI have the following discret ARMA (1,1)-Process for the Log-Price of electricity P:P(t) = a + b*[P(t-1) – a] + c*z(t-1) + z(t) a, b, c are constants; b for the AR(1) and c for the MA(1)-part respectivly t is the time index z(t) is normally distributed with zero me...
by Rez
July 31st, 2008, 8:18 am
Forum: Technical Forum
Topic: Granger Causality Test
Replies: 9
Views: 54230

Granger Causality Test

<t>"Granger-causality" is not "causality" in the everyday usage of the world.It merely means that lagged values of a variable help forecast another. Nothing more than that.So you can have X Granger-causing Y, and Y Granger-causing X or Y not Granger-causing XYou can have W causing X and Y, with X Gr...
by Rez
July 29th, 2008, 7:01 am
Forum: Technical Forum
Topic: mean reversion model for fx
Replies: 3
Views: 52843

mean reversion model for fx

depends on the currency pair.afaik for most currencies linear mean reversion is rejected empirically. The dynamics are a lot more complex than that.anyway, if you want to price a forward deal you need the yield curves, no?K
by Rez
July 24th, 2008, 9:42 am
Forum: Technical Forum
Topic: Equity L/S perfomance measure
Replies: 7
Views: 52193

Equity L/S perfomance measure

QuoteOriginally posted by: acastaldoProfit/Net exposure is what the investors care about.But if you don't like that you could take Profit/Gross Exposure i.e. 1/160 = 0.625%Well say I buy a forward instead, or any unfunded position. What is the ratio then?K
by Rez
July 17th, 2008, 8:03 am
Forum: Technical Forum
Topic: Backing out implied-volatility model
Replies: 3
Views: 52077

Backing out implied-volatility model

well you can write is as sigma_t = sigma_0 + c log S(0) - c log S(t) which impliesd sigma_t = - c * d log S_tand as log S_t follows the Brownian motion, I think the correlation would be indeed -1Kyriakos
by Rez
July 14th, 2008, 7:13 am
Forum: Technical Forum
Topic: Svensson Model - Euler discretization
Replies: 2
Views: 51676

Svensson Model - Euler discretization

The Svensson model does not assume any dynamics, just fits the yield curve. For that reason you can match with deterministic rates.Unless you recalibrate every day and consider the dynamics of the constants z1..z6Kyriakos
by Rez
June 17th, 2008, 2:19 pm
Forum: Technical Forum
Topic: CIR++ calibration
Replies: 5
Views: 56393

CIR++ calibration

<t>i don't think that they can all be calibrated together, as the model is not identifiedit would be best to minimize the errors of the CIR without the shift, then add the shift and see what happensor add a smoothness criterion of a minimum shift criterion on phi if you want to have everything toget...
by Rez
June 17th, 2008, 12:54 pm
Forum: Technical Forum
Topic: CIR++ calibration
Replies: 5
Views: 56393

CIR++ calibration

what are you calibrating? i mean what is your minimization function?the ++ part of the CIR++ should calibrate automatically to the CIR modelKyriakos
by Rez
June 5th, 2008, 2:45 pm
Forum: Technical Forum
Topic: eXTREME VALUE THEORY
Replies: 6
Views: 54925

eXTREME VALUE THEORY

These lecture notes have a couple of slides on tail index estimation and examplesThe very basics though, nothing too complexHope that this helpsKyriakos
by Rez
June 5th, 2008, 2:42 pm
Forum: Technical Forum
Topic: eXTREME VALUE THEORY
Replies: 6
Views: 54925

eXTREME VALUE THEORY

i guess for stock risk with EVT you might want to check the tail index
by Rez
June 3rd, 2008, 7:14 am
Forum: Technical Forum
Topic: How to identify position in correlation?
Replies: 5
Views: 55314

How to identify position in correlation?

I believe that if you buy (long) an equity tranche or sell (short) a senior tranche, then your position's value will increase with correlation. This is then a long correlation position.For mezzanine tranches it all depends on the attachment/detachment points.Kyriakos