How to remove the outliers in distribution fitter

3 views (last 30 days)
Raj Arora
Raj Arora on 8 Oct 2021
Commented: Jeff Miller on 9 Oct 2021
As shown in the dist. fitting there is some outliers in the starting and at the end is there any way to remove it?

Answers (1)

Jeff Miller
Jeff Miller on 9 Oct 2021
If you click on the "Exclude" button in the picture you posted, you will get the option to set lower and upper limits to exclude values outside certain ranges.
Alternatively, when you specified the dataset initially, you had the option to use a "censor" vector to exclude any data values you want.
I must say, though, that those extreme scores don't look like outliers to me--they look too regular. Instead, my guess is that the true distribution isn't normal but rather some distribution with heavier tails--maybe a Weibull, for example. So, I would recommend trying fits of all the data to some of the other distribution options
  2 Comments
Jeff Miller
Jeff Miller on 9 Oct 2021
Purely on a visual basis, the distribution looks almost like a gamma to me--except that the minimum has been shifted up from zero to about 3. So, if you subtract a constant from all of your scores so that they start just above zero, you might get a better fit to one of the built-in distributions for the new adjusted scores. I don't know whether that adjustment makes any sense in terms of how the scores were generated/measured, though.
I think you are always going to have a problem at the top end though. That abrupt drop-off at 16 looks like the measurement process imposed some kind of artificial upper limit.

Sign in to comment.

Tags

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!