Hi all,I am tring to implement the following idea but couldn't find any reasonable way to do. So could you please adive me on this matter??Set 3 risk parameters(Say 10yrPar for Yield curve level and 2-10Spread for frond end curve, 10-20 Spread for long end curve) to quantify yield curve risk. Then I want to do relative value analysis based on these 3 risk factors, meaning find securities which is traded rich/cheap compared with others(these rich/cheap is neutral to 3 risk factors)In other words, given 10yr Par rate and 2-10,10-20 spreads as fair, I want to find relative value in other maturity zone, say 7yr etc.... Would appreciate any help!Thank you very much,ksdt

That is tricky, because your going to have to assume (a more or less arbitrary) shape of the curve. I.e. how the curve looks between the 10 year and the 2 year, say linear. So you could easily get the relative value of the 7 year but just adjusting it by Yield(2)+((Yield(10)-Yield(2))/(10-2))*(7-2). This is obviously completely naive and needless to say will cost you your ass.It'll be very tricky to do relative value analysis based on a model like this. You don't like to strap a curve and then look at the individual points (or, heaven forbid: PCA). That way you could simply calculate all the spreads and determine their relative value this way.Just an idea.

Hi FDAX,do you mean that is useless to incorporate the knowledge about risk factors from PCA into a yield curve model (Svensson style)to derive what a fair curve is? based upon that fair curve could we gauge the richness of each sector or bond?

Derivababy, there was no statement by me about incorporating risk factors from PCA. The PCA eigenvectors point to the same facts as does simple parallel, tilt and curvature point of view. I don't see why someone would want to trade the intuitiveness of spreads, butterflies and boxes in favor of some strange vectors which almost are simple geometric shapes but not quite.Just my DEM 0.02