I propose enhancing the ALDEx2 glm function to support robust regression using M-estimators. The current documentation acknowledges that "large effects driven by outliers can result in false positives." This issue is particularly prevalent in microbiota taxa data, which often includes outliers that can skew results. While data modification methods like winsorization can mitigate this, they also result in data loss. Implementing a robust regression method, similar to glmrob from the robustbase package, would allow for parameter estimation that minimizes the influence of outliers without altering the original data.
I propose enhancing the ALDEx2
glmfunction to support robust regression using M-estimators. The current documentation acknowledges that "large effects driven by outliers can result in false positives." This issue is particularly prevalent in microbiota taxa data, which often includes outliers that can skew results. While data modification methods like winsorization can mitigate this, they also result in data loss. Implementing a robust regression method, similar toglmrobfrom the robustbase package, would allow for parameter estimation that minimizes the influence of outliers without altering the original data.