Skip to content

Subjective Outliers Removal

Subjective outliers removal is defined by Bakker et al., 20141 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 built-in 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.

  1. 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. ↩

Last update: 2021-09-18