joe

Thanks for your thoughts. Let me attempt to summarize and reply to your
concerns:

1) EG has provided an experience that did not meet your initial
expectations. In fact, you have been underwhelmed relative to your
expectations.

2) EG takes a long time to run. EG is mainly a front-end to your SAS server,
latency is typically experienced for server start-up, large log files to
return and large output datasets to be returned. Also, upon closing projects with many datasets, there is a lag as EG determines how to clean up after the project.

That said, EG is not any slower than local or server SAS via direct
connection. Often people attribute slowness to EG that is frequently related to growth in users with the addition of EG or server processes that, of which they are unaware.

3) EG is an insult to experienced SAS programmers. I would strongly disagree with you. In fact, I would be happy to stack my SAS programming skills head-to-head with anyone in the SAS programming world. I have worked in many companies using SAS for hand-coded data warehouses, innumerable analytic data prep programs, simulation, non-standard statistical analysis (not available in PROC’s), macro library development, statistical programming, complex SAS GRAPHs, forecasting, old-school AF and FSP development and many other areas.

I have rummaged through almost every SAS v 6 and v 8 SAS manual at some point to solve various problems. I have personally spent time training at least 50-60 SAS programmers over the years and I understand the time needed to become proficient and expert at SAS programming, even in a small fraction of SAS traditional capabilities.

I only state all of this to show that I am a very experienced programmer and I have also taken the time to master EG. I can state emphatically that EG can replace 80% of the code I have ever written and in much less time than I could write it.

I once met with an author of many SAS books and he said something similar to your statement. This is a person for whom I have great respect. I respectfully told them they were wrong and I wanted the chance to show them.

They pulled one of their books off the shelf and opened to a problem in the
middle, stating “EG can’t do this!” I asked how long it would take them, as
the author of this book, to write the code at the moment. They said around 5 minutes. I then asked for the data and opened EG. I solved the problem in the book in around 2 minutes.

They were incredulous and opened another example from another book. Same question, they stated they could do it in 3 or 4 minutes. I solved it about 2 minutes with EG.

Having said all of this, there are many times that I work with EG projects to
radically clean them up- both in verbosity and in efficiency. However, I was
able to explore multiple analysis paths with EG in a much shorter time than
writing all code from scratch.

4) “If you make a tiny mistake and it runs infinite, you cannot stop it with
the stop button anywhere. You have to abort it with task manger and try it
new.”

This is a weakness, but one of how the 1st generation SAS client-server
architecture behaves, but appears to users as an EG shortcoming. This has been addressed in SAS 9.2 and was somewhat better in EG 4.1/SAS 9.1.3.

5) “If you missed out saving it, you even need to remember the whole code.”

I agree; I save religiously using simple version numbers and cleaning up
afterwards. EG 4.2 has a new auto-save project feature.

6) “I would not mind its exisence. It’s great for non programmers.”

I agree, in fact this is why it was initially created. Why should someone
need to learn SAS programming to create a report, run a regression, etc.? In fact, this has been a big savior for SAS in my many visits to SAS customers. To be blunt, SAS wouldn’t be in 20-30% of all companies without it and I think this is a low estimate. Strong SAS programmers, while very valuable, are very hard to find and are typically bored working in a business group. They need a central team of comrades to be happy and have variety in their work. EG often allows these central teams to even exist and to offload mundane projects into the business.

7) “They just need to click click and they can call themself as programmers.
But then the experienced programmer needs to fix their mess they made.”

I have yet to meet an EG jockey that thought they were a “programmer”. They usually call themselves an “analyst” or something similar. That said, I think solid, personable programmers could immensely help heavy EG users with learning good data management practices, understanding how macros could be wrapped around EG projects, etc. You as the expert programmer are presumably an expert to enable business success, spending 10-15% of your time educating the masses of casual users will likely yield very large returns in my experience.

Also, many EG analysts would be horrible programmers, but they are often much better analysts than many SAS programmers in my experience. Why? Because they understand the subject matter of the data being analyzed and they can communicate the results in an actionable manner to the business owners. This is why you see very few “SAS Programmer” jobs in lines of business, because they are very focused on the details necessary for strong programming and not focused enough on the business needs and finding rapid answers to them.

Bundling it wit Enterprise Miner is just a taktic like buy one get 2.
Nobody I work with is actually using it.

I know the people who manage EM development and the rationale for adding EG was nothing like your statement. It really boiled down to the need for analytic data prep, plain and simple. If you have EM, you can totally avoid EG with Code Nodes if you prefer, but I would say you are overlooking a valuable tool for rapidly building analytic datasets. Also, there are many GRAPH and STAT PROCs that are very complex to write that EG can save you tremendous amounts of time to create, just like the EM point and click. I have used EM combined with EG at several companies and think it is the best data mining toolkit anywhere, if money is no object.

All the best,
Stephen McDaniel

Bundling it wit Enterprise Miner is just a taktic like buy one get 2. Nobody I work with is actually using it.