any guidance on how to prove the linearity of quadratic variation would be greatly appreciated.

Denote quadratic variation of a continuous stochastic process $X_t$ as $$\left[X,X\right]_t = p-\lim_{max_{i}|t_{i+1}-t_i| \to 0}\sum_{i=0}^{n-1}{|X_{t_{i+1}}-X_{t_i}|}^2$$ and given that the quadratic covariation of two continuous square integrable martingales $X_t$ and $Y_t$ is defined as \begin{align} \left[X,Y\right]_t &= \frac{1}{4}\bigl(\left[X+Y,X+Y\right]_t - \left[X-Y,X-Y\right]_t\bigr) \\ &=\frac{1}{2}\bigl(\left[X+Y,X+Y\right]_t-\left[X,X\right]_t - \left[Y,Y\right]_t\bigr) \end{align}

show that $$\left[aX+bY,Z\right]_t = a\left[X,Z\right]_t + b\left[Y,Z\right]_t$$

EDIT: I am mainly hesitant because of the modulus and square exponent. Perhaps, let's try to show a simplier example: $[X+Z,Z] = [X,Z]+[Z,Z]$ (I drop off the index)

LHS: $$[X+Z,Z]=\frac{1}{2}[X+2Z,X+2Z]-\frac{1}{2}[X+Z,X+Z]-\frac{1}{2}[Z,Z]$$

RHS: \begin{align} [X,Z]+[Z,Z] &= \frac{1}{2}([X+Z,X+Z]-[X,X]-[Z,Z]) + [Z,Z] \\ &= \frac{1}{2}[X+Z,X+Z] - \frac{1}{2}[X,X] + \frac{1}{2}[Z,Z] \end{align}

Should I be approaching this problem as such?


  • $\begingroup$ Denote $Q$ the quadratic variation. I think you should first prove that $Q(X+Y)+Q(X-Y) = 2Q(X)+2Q(Y)$. $\endgroup$ – Jakobian Jul 25 at 19:08
  • $\begingroup$ @Jakobian I don't quite see how the above property helps? $\endgroup$ – Galvin Ng Jul 27 at 4:44
  • $\begingroup$ What have you tried? It's really not that difficult, just a matter of using the very definitions of the objects appearing in the questin. $\endgroup$ – saz Jul 27 at 13:19
  • $\begingroup$ Well, I've thought about this in the following way. Quadratic variation looks a bit like a squared norm, so why not prove the rectangle equality to prove it's given by a scalar product? $\endgroup$ – Jakobian Jul 27 at 16:23
  • $\begingroup$ @Jakobian would you like to take a look at the answer I put up? Would you agree with my approach? I realized my initial doubt of using the triangle inequality was a mere scare. Just a tedious computation... Thanks! $\endgroup$ – Galvin Ng Jul 28 at 18:14

One approach would be to first show $$[f,g]_t = \lim_{\delta_n \to \,0}\sum_{i=0}^{n-1}(f_{t_{i+1}} - f_{t_{i}})(g_{t_{i+1}} - g_{t_{i}})$$ then we get$$[\alpha f+ \beta g,h]_t = \lim_{\delta_n \to \,0}\sum_{i=0}^{n-1}\left[(\alpha f + \beta g)_{t_{i+1}} - (\alpha f + \beta g)_{t_{i}})(h_{t_{i+1}} - h_{t_{i}})\right]$$ and $$\alpha[f,h]_t = \alpha\left[\lim_{\delta_n \to \,0}\sum_{i=0}^{n-1}(f_{t_{i+1}} - f_{t_{i}})(h_{t_{i+1}} - h_{t_{i}})\right]$$ $$\beta[g,h]_t = \beta\left[\lim_{\delta_n \to \,0}\sum_{i=0}^{n-1}(g_{t_{i+1}} - g_{t_{i}})(h_{t_{i+1}} - h_{t_{i}})\right]$$

Finally, we obtain the desired linearity of quadratic variation after a couple more lines of simplification.

  • $\begingroup$ "one approach would be to show [...]" What exactly do you want to show there? You are just using the linearity of sums and limits. $\endgroup$ – saz Jul 29 at 17:39
  • $\begingroup$ I am ultimately trying to show the linearity of quadratic variation. I understand that the first line is also a definition of the covariation in some texts, but it was not for the one I am using. $\endgroup$ – Galvin Ng Jul 29 at 17:48
  • $\begingroup$ It's an immediate consequence of your definition of the covariation $$[X,Y]_t = \frac{1}{4} ([X+Y,X+Y]_t- [X-Y,X-Y]_t).$$ Use the definition of the quadratic variation (for the right-hand side) and expand the squares. $\endgroup$ – saz Jul 29 at 18:50
  • $\begingroup$ Yes. I did expand the terms for $[\alpha X + \beta Y, Z]$ and obtained $\lim \sum \bigl[(\alpha X_{t_{i+1}} + \beta Y_{t_{i+1}}) - (\alpha X_{t_{i}} + \beta Y_{t_{i}})\bigr](Z_{t_{i+1}} - Z_{t_i})$ but I realized that is similar to the first line of the answer I posted. As such, I decided to start off (in my answer) as shown above. $\endgroup$ – Galvin Ng Jul 30 at 5:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.