Trends and Outliers - All Comments

re: Best Thing I Read Today

Tara Scherder — Wed, 29 Oct 2008 22:11:02 GMT

I love that statement! I've simply referred to 'dangers' of aggregation - like squelching important relationhips and Simpson's paradox. Avinash Kaushik gives it new dimension!

re: Spreadsheets Don't Cause Problems?

Scott Kravatz — Mon, 27 Oct 2008 15:00:18 GMT

Sounds more like a procedural problem than an Excel problem. To leave data in hidden cells on a spreadsheet is pure buffoonery. If they entered the numbers wrong would the problem be an Excel one? Garbage in, garbage out.

re: What Does Increased Integration of R Mean?

Michael Zeller — Thu, 02 Oct 2008 16:37:27 GMT

Tim, I definitely agree that predictive analytics is rapidly gaining momentum and will play an more and more important role in Business Intelligence.

Once you have a comprehensive view of your data, the natural next step is to look ahead (predict) and operationalize your knowledge, i.e., use predictive models to generate decisions in real-time. I look forward to seeing the integration of Spotfire and S+ compete in the market!

Other than a "wholesale integration" of R, there is the option to exchange models between different software packages using the Predictive Markup Modeling Language (PMML) standard promoted by the DMG (http://www.dmg.org). Once data analysis is complete and you would simply like to use the predictive model on new data, PMML is an excellent route to deploy your predictive models in a scoring engine,

For more about how R supports the export of PMML, please see:

adapasupport.zementis.com/.../how-can-i-export-pmml-code-from-r.html

re: Quick Reactions

Jeff (no, the other one) — Fri, 15 Aug 2008 12:10:39 GMT

"Fresh, real-time data...? What are ya, some kinda troublemaker? How can I justify making decisions the traditional way -- with my dartboard, of course -- when there's real, valid data all the time sucking the fun out of it?!?"

Read about you folks in CIO (The Price of Gas, June 2008). I like what I see. This is the kind of thing we've wanted to do for... ever. Thanks!

Jeff Jones

Dept. of Postsecondary Education

re: Everyone a Data Analyst

Bill Carney — Thu, 17 Jul 2008 18:54:52 GMT

While I may agree that having "every employee as an analyst" generally helps, only certain folks within each organization can even access the data and let alone can look at it properly. People removed from the process need guidance and help along the path – The same way you build a storyline in slides. From a presentation POV using the data as facts and how they are interpreted shortens the discussions considerably. An old boss of mine used to say "facts are friendly" I've modified my thinking less on the friendly part and more on the facts are facts, data is data, and numbers are numbers. Enterprise adoption and a top down commitment would be great BUT it may be difficult to start that way. Departmental presentations using data coming from accepted data sources (CRM, Data Warehouses, and Accounting Platforms) to propose plans of action seems to me a good way to start selling the process internally to the senior staff.