[SystemSafety] Collected stopgap measures

[Derek M Jones]

Paul,

> Just because Boehm's works are rather large, and widely cited, doesn't 
> make them correct.

It shows the power of promoting one's work, which Boehm does heavily.

> Laurent Bossavit went all the way down to the bottom a couple of the 
> rabbit holes in [1], and from that research Boehm was apparently just as 
> prone to fallacious thinking as the rest of us.
> 
> [1] https://leanpub.com/leprechauns

This is a great book.

Boehm's original analysis was based on a sample of 63 projects.
Large when he wrote his book, but small by today's standards.

He over-fits this data and quotes accuracy to two decimal places,
while the data has a standard error at the level of one-digit
accuracy.

There is also mistake in the fitting of one of his models:
https://shape-of-code.coding-guidelines.com/2016/05/19/cocomo-how-not-to-fit-a-model-to-data/

I find this very surprising, given the data has been public for almost
40 years.  Why has nobody thought to check it and raise the issue?


-- 
Derek M. Jones           Software analysis
tel: +44 (0)1252 520667  blog:shape-of-code.coding-guidelines.com