Response distribution

Are your response variables distributed...


 






...linearly
    
 







...unimodally
 








Or will you work with (dis)similarity matrices of your response variables?