Subjective Outliers Removal¶
Subjective outliers removal is defined by Bakker et al., 2014^{1} as a special and even more questionable case of outliers removal where the researcher lowers k (the threshold of considering an item to be an outlier) dynamically during the data analysis.
While we are able to mimic this behavior by tweaking the parameters of outliers removal, SAM has a builtin hacking strategy for this specific method.
!!! hackingstrategy “Subjective Outliers Removal”
{
"name": "SubjectiveOutlierRemoval",
"min_observations": 5,
"range": [
2,
3
],
"step_size": 0.5,
"stopping_condition": [
"sig"
]
}
The main difference between this algorithm, and general outliers removal is that fact that SAM generates an array of equally distant k’s between the given range
using the given step_size
. For instance, in the above example, a range of k’s will be {3, 2.5, 2}. Notice that the algorithm starts from highest to the lowest k.

Marjan Bakker and Jelte M. Wicherts. Outlier removal, sum scores, and the inflation of the type i error rate in independent samples t tests: the power of alternatives and recommendations. Psychological Methods, 193:409–427, 2014. URL: https://doi.org/10.1037%2Fmet0000014, doi:10.1037/met0000014. ↩