`isoutlier`

Detects outliers in data.

Category: Data Import and Analysis > Preprocessing Data

Syntax

TF = isoutlier(A)

true for elements more than three scaled median absolute deviations from the median.

TF = isoutlier(A, method)

Uses another rule, such as 'quartiles' or 'mean'.

[TF, L, U, C] = isoutlier(___)

Also returns the lower and upper thresholds and the center value.

Description

With the default method the scaled median absolute deviation is 1.4826 * median(abs(A - median(A))). A point farther than three of these from the median is flagged. Because median and MAD barely move when a few extreme values are present, the rule is robust, unlike the mean and standard deviation.

The result tells you which elements to inspect; use filloutliers to replace them or rmoutliers to drop them.

Input arguments

A: Numeric vector or matrix (columns are analyzed).
method: 'median' (default), 'mean', 'quartiles', 'grubbs' or 'gesd'.

Output arguments

TF: Logical array, true at outliers.
L, U: Lower and upper thresholds.
C: Center value (the median for the default method).

Example

Spot the odd value

x = [1 2 3 4 5 6 7 8 9 10 50];
[tf, lo, hi, c] = isoutlier(x)
x(tf)

Output

tf = 1×11 logical array

   0   0   0   0   0   0   0   0   0   0   1
lo = -7.3434
hi = 19.3434
c = 6
ans = 50

isoutlier