Normally distributed data errors

1. Sep 9, 2011

James.L

1. The problem statement, all variables and given/known data
Hello

If my data errors are normally distributed, is this the same as the data being normally distributed? I mean, by "normally distributed data errors" is meant that with 68% confidence the data lies within the true value?

2. Sep 11, 2011

Stephen Tashi

Re: Statistics

You aren't expressing your question clearly. It isn't clear whether you have a set of measurements (such the heights of various persons) or a set of vectors of measurements (such as the height and weight of various persons). So it isn't clear what a "data error" would be.

For example, if we use the rather strange terminology that the average height h_bar of the population is the "true" height, then for a person's height h, we could call h - h_bar an "error". However, I don't think most people regard deviating from the average as an error.

If we have an equation H(.) that attempts to predict a persons height from their weight then for a given person with weight w and height h, we could call the quantity (h - H(w)) an "error" since it is an error in in how the equation predicts the weight.

Last edited: Sep 11, 2011
3. Sep 12, 2011

James.L

Re: Statistics

I think I get it. My original question got answered by your post, even though my question is poorly formulated.

Thanks!