More of a stats question. I have a non normal distribution of data. The data seems to fit the 10th to 90th percentile well while leaving out the outliers. I decided to bootstrap and got a tighter interval (n is decent 48) for a 95 percent CI. Should I use the parametric CI? Does bootstrapping 'fix' non normalcy? Thoughts? Pitfalls? I can delve deeper if need be. Thx guys and gals.
Asked
Active
Viewed 150 times
0
-
Interesting question, but since it's a discussion question it's off topic for SO; try stats.stackexchange.com instead. That said, I don't recommend omitting outliers. Outliers just mean your data are all over the place -- if you omit them, you are essentially claiming the process is cleaner than it really is. My advice is to just let it all hang out -- if your intervals are all over the map, that's just the way it is. – Robert Dodier Dec 15 '20 at 04:01
-
Thank you Robert! I will check there. I do concur about the outliers, but in this dataset I have decent reasons to believe they have a very low (but obviously not none) chance of occurring. Its an interesting problem feeding the rest of a model. If I trim the outliers my monthly prediction for use gets better but the yearly prediction gets worse and vice versa. So if I could get a bootstrapped mean out of the non normal distribution how well does a 95% parametric CI work and how many cardinal sins did I commit? Or does it just come down to doing it and testing the results? I'll read some more. – S. Pitcher Dec 16 '20 at 02:35