fix log_likelihood function

[tokoharu]

There is wrong statement in the function log_likelihood.

Specifically, there is an error in the calculation of the normal distribution, which I would like to correct with this request.

[pfmonville]

Hi,
I wonder if the integer division is really what’s wanted here.
In method log_likelihood line 26 with this part -(rd**2)/2/sigma2[i] since rd is an integer when I go through your test case I obtain 6.328759366440833e-15 instead of 1.2091379396665127e-07 the first time it hits line 26.
(at first rd is -1 so:
(-1)²=1 -> 1/2=0 instead of 0.5)
Is this the desired behaviour ?
If this is the calculation of the normal distribution I think it should be a floating point division.

With floating point division I got a different result for the first day, then it converges.
yours -> mine
-102.65265035957913 -> -69.65648196168772
-1.9397850625546684 -> -1.9397850625546684
-2.1269280110429727 -> -2.1269280110429727
-0.6931471805599453 -> -0.6931471805599453

[tokoharu]

@pfmonville
Thank you for pointing it out!!
It's not desired behaviour.
I'll fix it later, so please wait.

[tokoharu]

@pfmonville
I confirmed that, in the latest commit, result for the test is same as your result.
I'm not good at Ruby, so, thank you again for pointing out my mistake.