Hi,
I have matrix 'x' (see pic) with 5 values. Out of those one value deviates from the rest. So when I take the mean, standard deviation and plot with errorbar in Matlab, the deviation seems to be more. Is there a way or script or function to automatically neglect values which are away from the group and then proceed with the mean ?
I would want this to detect automatically
Thanks,
Praveen
Instead of throwing away the value, have you considered other possibilities that it may be a real but more rare event. Being present in 1 out of the 5 values, this perhaps is not that rare as I suggest earlier. You may want to use a likelihood based approach to calculate the mean and using a t distribution with a heavy tail - that should possibly avoid the mean being affected by the outlier.
try this https://uk.mathworks.com/help/stats/fitdist.html?requestedDomain=true it may be useful or point you in the right direction.
Also you can define a derivation function and set a limit for it. If the value was greater than the defined threshold, it should be deleted.
This can be done in a number of ways, such as setting thresholds or setting limits on derivative values, as described in previous answers.
However, I hope you have properly followed a scientific method for event observation. 1 out of five is 20%! Try more points and/or repeating the experiments. If the values is consistent, try exploring the literature and understanding the phenomenon causing it. This 'one point' we usually smother could be the best part of our research.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics