I'm hoping that someone can help me understand why someone would call something an RMSD that is really just a squared residual. I'm using someone else's R script which performs simulations according to a model. The model has a free parameter that needs to be optimized, so a method is needed to determine what value for this free parameter results in a better fit. To optimize the value, two different values are compared using what the author refers to as an RMSD, which they calculate as follows:

rmsd <- (model_mean - observed_mean)^2

To determine which value for the free parameter results in a better fit, an "rmsd" is calculated for both simulations, and whichever "rmsd" is closer to zero is the better fit. This seems like a fine way to determine which of two values is better for the free parameter, I'm just puzzled because it's not an RMSD, as far as I can tell. I understand RMSD to be calculated as follows:

rmsd <- sqrt(
          sum((model_mean - observed mean)^2) / n

This is different since it's the square root of the mean of numerous squared residuals, so I guess the only difference between this and what the author calls RMSD is that they don't get the square root. Am I missing something, or am I right in being confused about why they call a squared residual an RMSD?

Sincerely, Statistically ignorant


1 Answer 1


The "r" might stand for "residual", and there seems to be a "mean" in the calculation, so the term "rmsd" may make sense as a "residual mean squared deviation". I agree that it's not a very good abbreviation.

I don't think you will get a much better answer here, which would require reading the mind of the original script's owner. Can you possibly get a hold of them?


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.