Go forward by going back

Seeking a better understanding.

Posts Tagged ‘data

Using data for effective work

leave a comment »

By a system I mean that the diverse people actually work together to direct their specialized capabilities toward common goals for patients. They are coordinated by design. They are pit crews. To function this way, however, you must cultivate certain skills which are uncommon in practice and not often taught.

For one, you must acquire an ability to recognize when you’ve succeeded and when you’ve failed for patients. People in effective systems become interested in data. They put effort and resources into collecting them, refining them, understanding what they say about their performance.

Second, you must grow an ability to devise solutions for the system problems that data and experience uncover…

That’s from Atul Gawande’s commencement speech at Harvard Medical School, entitled “Cowboys and Pit Crews.”

While Dr. Gawande is speaking directly to physicians, I think that the paragraphs above apply equally well to all professions. The idea is to admit that you can’t you can’t do everything by yourself, to recognize the strengths of others, to promote diversity in your life and work, and to use data to help you understand what you are doing right and what you are doing wrong.

Read the entire speech.

Written by Niall

May 27, 2011 at 6:29 pm

Posted in Data

Tagged with , ,

Off-equilibrium correlations and Eyjafjallajökull

leave a comment »

Volcanoes and Iceland are back in the news. so let’s follow that theme and play with Google Labs new tool, Google Correlate.

Give it a test run. I suggest searching for “volcano”, then selecting “iceland” (Direct link, here). Looking at this data is a great example of why choosing your time period matter when performing analysis.

For the total period, 05 January 2003 to 24 April 2011, correlation of search activity for iceland and volcano was 0.8859. “Wow, that’s high!” You may say. Indeed, you’d be right.

Now, if we divide the period into one period before and one period after 20 March 2010 (the date of the first eruption of Eyjafjallajökull eruption in 2010), we can find a remarkably different picture. In the first period, “iceland” and “volcano” show measly correlation of 0.5563. After the first eruption, the two search terms seemingly move in unison with a staggering correlation of 0.9811.

Not so fast. Let’s step back and look at this graph. If  we concentrate on the portion of the graph where most of the scatter plot is concentrated, it is hard to see any relationship between the two search terms. If you put your hand on your computer screen to cover the two left-most dots, the blue trendline is no longer likely to be upward sloping. In fact, if we take out those two right-most weeks of search data the correlation falls to 0.7893. And if we look at the correlation, from 16 May until the latest date, the correlation returns to it’s pre-Eyjafjallajökull level (rho=0.5912).

Kudos to Google for making such an excellent tool and resource publicly available.

Written by Niall

May 25, 2011 at 8:18 pm

Posted in Uncategorized

Tagged with , , , , ,