Take THAT, PHBs of the world...
Finally finished up a side project I've been working on. In 1999, a Business dissertation from Georgia Tech described a really well done study on the risk factors for software engineering projects. It covered a wide gamut of issues, including things like user buy-in, upper management support, team dynamics, turnover, complexity of the application, etc. *Extremely* well done study that had a nicely large sample pool.
Only one problem.
Project complexity correlated the lowest of any of the factors to project success. .29 to be exact.
This was... stunning. Not to mention utterly at odds with everything from my own experiences and most engineering oriented literature produced in the last 30 years.
My advisor responded with "Well what do you expect? They're a bunch of management wonks who don't get the technical side." While true, I felt that if there truly was a correlation at all (which there should be), that it would come out in the final data regardless of the respondents' *intent*.
So I re-analyzed it from an engineer's point of view, and regrouped the data according to technical guidelines as well as the business oriented divisions already in the paper, and then a second-order grouping was formed off of the initial technical ones.
New correlation? 0.86.
*And the model fits the data more precisely.*
Yay me.
Only one problem.
Project complexity correlated the lowest of any of the factors to project success. .29 to be exact.
This was... stunning. Not to mention utterly at odds with everything from my own experiences and most engineering oriented literature produced in the last 30 years.
My advisor responded with "Well what do you expect? They're a bunch of management wonks who don't get the technical side." While true, I felt that if there truly was a correlation at all (which there should be), that it would come out in the final data regardless of the respondents' *intent*.
So I re-analyzed it from an engineer's point of view, and regrouped the data according to technical guidelines as well as the business oriented divisions already in the paper, and then a second-order grouping was formed off of the initial technical ones.
New correlation? 0.86.
*And the model fits the data more precisely.*
Yay me.