Error detected by tests of LogNormalFunctionTest class

Issue #143 new
Manuel Luque created an issue

Fix error found by tests of class LogNormalFunctionTest.

Comments (6)

  1. Francisco Javier Díez

    In the network attached, the uncertainty for HealthState(Disease=yes,Therapy=no) is modeled by a logNormal with mu=ln(3)=1.09861228867 and sigma=1. Therefore, the value shown in the cell should be exp(mu) = 3, but it is 4.946164.

  2. Manuel Luque reporter

    The mean of LogNormal(mu, sigma) is: exp (mu + pow (sigma, 2.0) / 2.0).

    Thus, for the introduced values the result is 3.0 x exp(1/2) = 3.0 x 1.6487212707 = 4.9461638121.

    A log normal with mean of 3 would be obtained if we also take the logarithm when introducing the value for sigma, i.e. log(1) = 0.

  3. Francisco Javier Díez

    You are right. I made a mistake when reading the Wikipedia. My apologies.

    Then my proposal is that, in addition to having a function logNormal, we implement a new function logNormal-mv, where "mv" stands for "parametrized with the mean and the variance". This is similar to the difference between the functions Gamma and Gamma-mv that we already have.

    Anyhow, I think we should rethink the idea of having different parametrizations for second-order probabilities.

  4. Log in to comment