lundi 26 juin 2017

determining standard deviation of normal distribution

My model can only have values between 0 and 1. There are 3 tresholds within the interval [0,1], namely:

  • 0.07 - low
  • 0.65 - middle
  • 0.95 - high

I would like to create 50 random values that are distributed normaly around each treshold. I know that for this the function rnorm() can be used. However I do not know the variance (there is no data that can be used to estimate it). I would like the variance to be proportional to the interval in between the values - 0.58 between low and middle, 0.30 between middle and high. For now i have done this:

se.sw <- rnorm(50, 0.07, 0.07/3)
se.m <- rnorm(50, 0.65, 0.58/3)
se.st <- rnorm(50, 0.95, 0.30/3)

I have chosen 3 with no specific reason. I am aware that the values may overlap - for example that value from middle is higher than value from high. that is OK, as long as it is not too much overlay. Ideally it would be around 10%.

What would be a sensible way to achieve this?

Thank you.




Aucun commentaire:

Enregistrer un commentaire