Changing Alpha value for ANOVA giving different features

[Jonathan Greenberg]

Hello,

I apologize for having a second topic, but in addition to my previous question, I have run into a discrepancy I cannot explain and was hoping to get your insight. 

I'm not sure if more information is required, but what happened was I ran a dataset through LEfSe with Alpha values of 0.05 and 0.01, and while the majority of the features from the 0.01 results (18 significant features) were also in the 0.05 results (54 significant features), 2-3 features were found in the 0.01 analysis but not the 0.05 analysis. I'm not sure how a feature can be significant at 0.01 but not at 0.05. 

This dataset consists of 3 biological sample types from 25 individuals and controls so the metadata included Class and subject ID. All parameters were ran as default except the Alpha change and that this was performed as One-against-one.

I'm not a statistician so I apologize if there is an obvious/easy explanation for this. I ran this analysis through both Galaxy and biobakery just to confirm the method was not the issue.

Thank you!

Jonathan Greenberg