A flexible method for detecting system-wise outliers is presented. This approach attempts to simplify
the detection of anomalous observations without relying on large datasets and complicated statistical
modeling techniques. The method utilizes the leave-one-out method, similar to cross-validation technique in
data resampling. Using this approach, observations are iteratively removed and a measure of variability is
recalculated in each iteration. By doing so, the influence of each observation is evaluated in the data. Results
of the simulation study show that this new method adequately recognizes anomalies in several data scenarios
while minimizing incorrect detections. The method worked particularly well in data scenarios with high number
of time points and low to moderate variability. On the contrary, the method showed weak performance when
applied to datasets with low number of time points and very high variability. The proposed method was also
applied to detect irregularities in regional fisheries production in the Philippines. Upon visual validation, results
show that the new method was able to detect irregularities in the production data, such as trend and extreme
shocks.