Measuring Usability: Ask a Statistics Question

October 23, 2007 | asked by Anonymous :

Question : How do you make data normally distributed?

Answer : Many statistical techniques require data to be roughly normally distributed. For example, if you want to compute confidence intervals or a margin of error on non-normal data the results will be inaccurate to the degree your data deviates from normality. If your data is not normal, say its highly skewed, such as time on task data, which is often positively skewed. There are generally two things you can do,

Transform your data so they are roughly normal. This involves taking the logarithm or reciprocal as two common tansformations on the data. You then use the transformed data to compute the confidence intervals and transform back the results into the original scale when reporting.
Use methods that don't require normality. For confidence intervals you can use re-sampling techniques such as Monte Carlo and bootstrap. For tests of significance there are non-parametric tests that use ranks.

How helpful was this answer?
Avg. Rating: 0 ( 0 )

Question Tags

Tag Name	#	Vote
Bootstrap	1
Log Transformation	1
Monte-Carlo	1
Non-Normal Data	1
Non-Parametric	1
Normal Distribution	1
Transformation	1

New Tag:

User Provided Answers & Comments :
No user comments or answers posted yet.

Comment or Update Answer:
Name
Email Address Not Published