Hi there,I'm working on the JY-inflation model as described in Mercurios inflation paper Mercurio's paper.In order to get prices for caplets and floorlets one has to calculate the variance of . There is a formula for the variance in Mercurios paper, page 15.To my mind, this expression should be positive by definition.BUT: For certain choices of the model's parameter set (alpha_n, sigma_n, alpha_r, sigma_r, sigma_I, rho_nI, rho_rI, rho_nr) the formula gives negative results, i.e. a negative variances.Can that be? I don't understand it... Any hint is highly appreciated...

Can you provide the parameters that gives negative results ? I can try them on my implementation of this formula.

Hi, here is an example: an 0.03, sn 0.01, ar 0.0446314, sr 0.0175392, si 0.0160384, rho(nr) 0.370196, rho (ri) 0.912015, rho (ni) -0.99999, t 0.906849. The variance you get would be slightly negative and this is attributed to the rho(ni) value.....

QuoteOriginally posted by: quantericHi, here is an example: an 0.03, sn 0.01, ar 0.0446314, sr 0.0175392, si 0.0160384, rho(nr) 0.370196, rho (ri) 0.912015, rho (ni) -0.99999, t 0.906849. The variance you get would be slightly negative and this is attributed to the rho(ni) value.....The negative variance comes from the fact that one of the eigenvalues of the correlation matrix is negative. The correlation matrix must be positive semidefinite.