The Henderson-Hasselbalch equation is given by

$$\text{pH} = \text{pK}_{\text{a}} + \log_{10} \left( \frac{[\text{Base}]}{[\text{Acid}]} \right)$$

where $\text{pH} = -\log [\text{H}^+]$. Solving for the $\text{pK}_{\text{a}}$ we obtain the equality

$$\text{pK}_{\text{a}} = \text{pH} - \log_{10} \left( \frac{[\text{Base}]}{[\text{Acid}]} \right).$$

Just to make our notation simpler, let us substitute:

  • $W :=\text{pK}_{\text{a}}$
  • $X := [\text{H}^+]$
  • $Y := [\text{Acid}]$
  • $Z := [\text{Base}]$

giving us the expression

$$W = - \log_{10}X - \log_{10} \left( \frac{Z}{Y} \right)$$

which is equivalent to

$$W = \log_{10} \left( \frac{Y}{XZ} \right).$$

Those familiar with using the equation may raise an eyebrow at treating $\text{pK}_{\text{a}}$ as a random variable since it is often taken to be a constant. Here I am supposing that variability enters into our calculation of the $\text{pK}_{\text{a}}$ through the variables $X,Y,Z$ that are themselves not known with perfect precision.

Assume that $X,Y,Z$ have a joint log-normal distribution. Statistical independence cannot be assumed.


What is the distribution of $W$?

  • $\begingroup$ The answer below is correct but I assume you are also looking for the detailed description of the mean and variance. If $\log X$, $\log Y$, and $\log Z$ follow a multivariate normal with the usual parameters, then $W$ is normal with mean $(\mu_Y-\mu_X-\mu_Z)/\log (10)$ and variance $(-2 \rho_{XY} \sigma_X \sigma_Y+2 \rho_{XZ} \sigma_X \sigma_Z-2 \rho_{YZ} \sigma_Y \sigma_Z+\sigma_X^2+\sigma_Y^2+\sigma_Z^2)/(\log{10})^2$. $\endgroup$
    – JimB
    May 26, 2022 at 6:20

1 Answer 1


$W$ is Normal:

$$W=-\log_{10} X -\log_{10} Z+\log_{10} Y$$ and by assumption the three quantities on the right-hand side are multivariate Normal, so their sum is Normal.

  • $\begingroup$ I am aware of the stability property of the normal distribution which requires independence. How is the result obtained in the dependent case? $\endgroup$
    – Galen
    May 26, 2022 at 6:06
  • $\begingroup$ These are multivariate Normal distributions -- that's what it means for $(X,Y,Z)$ to be multivariate logNormal. So any linear combination of them is also Normal $\endgroup$ May 26, 2022 at 6:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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