0

I want to weight the two variables based on the amount of error. More the error lesser the weight I want to give.

e.g My variables are the distance from the origin(0,0) to two points with a gaussian noise of (0 ,5m) .

     dist1 = 2.8m  , Var(dist1) = 13.05, Weight1 = 1/Var(dist1)  = 0.08

     dist2 = 70m ,Var(dist2)  =25.5. , Weight2 = 1/Var(dist2)  = 0.03

My problem is dist1 is getting more weight even its measurement is noisier than dist2 . What I can do correct this issue. Thanks for your time and help.

Note: In actual problem, I don't know the real distance, but i know the mean and variance.

vicky
  • 13
  • Why do you say dist1 is noisier? The variance $13.05$ is smaller than the $25.5$ of dist2. – Lee David Chung Lin Dec 11 '16 at 04:12
  • @Lee the variance give an idea of spread from the mean, so in my understanding, the max percentage errors in the samples of dist1 can be far more than the percentage error in distance2 . or i can say that the amount of error introduce is same in both dist1 and dist2. the dist2 is bigger than dist1 , so the percentage error will be more on the dist1. It's my understanding, may be i am entirely wrong. so please correct me. – vicky Dec 11 '16 at 08:02
  • yeah it's better to use the percentage error, just that your code didn't use percentage error as the weight. – Lee David Chung Lin Dec 11 '16 at 08:27

0 Answers0