Degrees of freedom in standard deviation computation

[daikitag]

When performing validation, I realized that there is a slight discrepancy between the phenotype simulation output of tstrait and AlphaSimR, and it was due to the degrees of freedom input when computing the standard deviation https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.std.html. AlphaSimR uses $N$ as a divisor instead of $N-1$, so we should also add one more input ddof to normalise_phenotypes function, such that the user can define how the standard deviation is being computed.

[gregorgorjanc]

FYI We do this because our AlphaSimR populations have all the data so we are not calculating sample variance, but population variance

[daikitag]

FYI We do this because our AlphaSimR populations have all the data so we are not calculating sample variance, but population variance

Thank you for letting me know about it. I digged inside the AlphaSimR code to find out where the discrepancy is coming from, and I'm glad to hear your explanation.

[gregorgorjanc]

Feel free to reach out for any AlphaSimR questions;)