I have a set of independent software development metrics such as: Deployment Interval (Range: 1yr to 1sec), Feature Cycle Time (Range: 3 mon to 1 day), Number of Features Released (Range: 0 to 100's). Opposing ends of these ranges represent low value at one end and higher value at the other. I'm trying to find an (likely multi-step) algorithm to combine these into a composite value between 0.000 and 1.000. I envision both linear and exponential transformation methods. Are there data transform libraries that can normalize differing data values, over differing range, using differing units of measure?
Hai,
The Composite metrics define a new higher-level metric by specifying an arbitrary set of mathematical transformations to perform on a selection of native metrics you're sending us. ... Alternatively you can think of it as a way to specify complex queries against your native metrics.
To create a composite score for each of the domains, we will have to combine the respective five items to create a score for each domain, and we can combine the 15 items to make an overall satisfaction score. In most cases, you should combine the scores according to the instructions from the instrument developers.
For more info:
Article A composite metric for assessing data on mortality and cause...
C K Gomathy Thanks so much for this paper. I've interpreted the methodology outlined as: 1) Create ratio of each performance dimension; 2) Run simulation on each dimension; 3) Check simulation against empirical data-to confirm weighting; 4) Compute Composite Index; 5) Smooth data (as needed)
I would simply use a Lagrange cost function. Tuning its lambda multipliers depends on your semantics
- Badges
- Science topic
- Similar topics
- Social Science
- Media