HomeScience & TechnologyRelated VideosMore From: SAS Software

SAS Enterprise Miner Tip: Imputing Missing Values

84 ratings | 36564 views Jeff Thompson, a statistical training specialist with SAS Education, provides an overview of the predictive modeling portion of the SAS training course "Applied Analytics Using SAS Enterprise Miner." Thompson also provides a tip on the imputation of missing values. To learn more about the SAS training course "Applied Analytics Using SAS Enterprise Miner," visit
Get embed code!
Text Comments (12)
Yannick Luxi (2 years ago)
Hi and thank for this video. Do you think that filter is always useful for deleting the lines that contain outliner or should we in sas use the replacement note ? Is there a video dedicated to this topic ? thanks you again
SAS Software (2 years ago)
Aside from the Enterprise Miner SAS Communities, we'd recommend taking instructor led training. If your interest is a detailed understanding of neural networks, SAS Education does offer an advanced course on Neural Networks in Enterprise Miner Note that you should be familiar with Enterprise Miner and predictive modeling in general before taking the class.
Yannick Luxi (2 years ago)
thanks for this comprehensive answer ! I am trying to understand the computation of the neural network right now and all the effects of the different parameter. It's not the easiest part. If you got some tips let me know :)
SAS Software (2 years ago)
Thank you for your patience! It depends if the goal of the analyst is to delete (remove) the cases or wants to keep the cases in but replace the value (of the outlier). To delete the observation altogether, use the filter node. To replace the outlier value with something else, use the replacement node. If the question is “should” I delete a case with an outlier, it depends. “Typically” an analyst does not remove or delete a case just because it has an unusual value, unless there is specific reason to do so (such as known bad/false data). Often the analyst may just leave it in as is! If there is a justified reason to keep the case in but mitigate the effect of the outlier, then the value can be replaced. Note, the reason we filtered out cases in the video, is that those were cases not relevant to the analysis. If you discover some records outside the scope of the analysis, they should be removed to prevent biasing the sample. We hope this information helps! Feel free to post programming/content related inquiries to our SAS Commiunites!
Yannick Luxi (2 years ago)
Thank a lot. My purpose is to check if there isn't outliners in my data, I don't know which note will suit the best filter or the replacement for this goal. I do not know if I had formulated it clearly :)
SAS Software (2 years ago)
Thanks for posting! We're checking on this for you!
Vangelis Dalucas (3 years ago)
Hi and thank you for this video! Is imputing suggested when clustering? I am at the moment setting the ''missing values'' option of the cluster node to ''mean'' (I suppose this is another way of imputing) but that does not seem to affect the segments greatly even when missing percentage is >>50% thank again
Vangelis Dalucas (3 years ago)
Thanks a lot!
SAS Software (3 years ago)
Hi Vangelis, Thanks for your comment! It is a best practice to impute in the presence of missing data.  Unless there is a business reason to OMIT cases with missing values, usually imputing is preferred when the percent of missingness is not too high.
txpoolstud (6 years ago)
Very nicely done...clear and articulate.  Appreciate it!
Anshul Srivastava (6 years ago)
hi...nice.thanks for this  video...however where is the link to take this course online??
SAS Software (6 years ago)
Hi Anshul! Glad you liked the video. Here is a link that will take you to our e-Learning page:  Hope this helps!