Reputation: 11
If amount of data points is small (i.e. 20 data points). Do I need to check its normality before calculating confidence interval?
Can anyone suggests a rigorous process to plot 95% confidence interval?
Thanks!
Upvotes: 1
Views: 374
Reputation: 731
With so few data points and if you don't know if the distribution is normal I would look into using bootstrap confidence interval. It's a non-parametric method so you are not assuming normality. The MATLAB function bootci
implements this method.
Here is the documentation for bootci
.
Upvotes: 2