You would probably want to determine first which type of suppression is present (i.e., classical, negative, reciprocal) and describe the results accordingly. Best, Christian Geiser
Thanks for your video! Very interesting! So, the suppressor effect is not a problem when we use the model to predict the y variable, right? I mean, in the model included x1 and x2, we can not say the verbal ability has a negative effect on the job_success, right?
Yes, that is correct. The sign of the regression coefficient isn't interpretable in this case. Nonetheless, the model with the suppressor included may be useful in terms of its predictive power; in fact, it is better than when the suppressor is excluded. Best, Christian Geiser
@@QuantFish I'm not understanding why the effect of the suppressor variable is uninterpretable. In your illustrated example, couldn't we say something like: the independent effect of verbal ability on job success is negative, controlling for the effect of exam scores on job success. In other words, when we take into account the fact that people with higher verbal ability tend to have higher exams scores (which are themselves positively associated with job success), higher verbal ability is actually associated with less success. Is it wrong to offer such an interpretation?
Thank you for the video! What are the next steps once a suppression effect ? How would we write this up in a results section for example?
You would probably want to determine first which type of suppression is present (i.e., classical, negative, reciprocal) and describe the results accordingly.
Best, Christian Geiser
Thanks for your video! Very interesting! So, the suppressor effect is not a problem when we use the model to predict the y variable, right? I mean, in the model included x1 and x2, we can not say the verbal ability has a negative effect on the job_success, right?
Yes, that is correct. The sign of the regression coefficient isn't interpretable in this case. Nonetheless, the model with the suppressor included may be useful in terms of its predictive power; in fact, it is better than when the suppressor is excluded.
Best, Christian Geiser
@@QuantFish I'm not understanding why the effect of the suppressor variable is uninterpretable. In your illustrated example, couldn't we say something like: the independent effect of verbal ability on job success is negative, controlling for the effect of exam scores on job success. In other words, when we take into account the fact that people with higher verbal ability tend to have higher exams scores (which are themselves positively associated with job success), higher verbal ability is actually associated with less success. Is it wrong to offer such an interpretation?