I have three group that will be performing code review on three standard C programs of K & L package. The length of these program varies from 50 LOC to 250 LOC. I will be measuring number of defects vs LOC and number of defects vs review time. Do my approach needs ANOVA technique?
Syed:
As you may know, ANOVA is a collection of statistical procedures used to explore and measure the differences among parametric statics associated with a given measurement activity.
If you need to use ANOVA proceudres depends on the objectives of your work.
There are many open source programs, R for example, that help with the calculations.
It is easy to do and may give some additional insight to your results.
Take care, be good to yourself and have fun,
Joe
I think you are going to have serious sample size problems. Let's say you use the model that there is a reviewer group effect (2 degrees of freedom) and a program reviewed effect (2 degrees of freedom). And then there is 1 df for the overall mean. That leaves just 4 degrees of freedom to support estimating 5 parameters. I think your confiedence intervals for your estimates are going to rather huge and almost surely uninformative. You really need a lot more data for this exercise to bear fruit.
I think you need to use Chi Square to test the relations between the variables in each group. And use ANOVA to determine any difference between the means of each variable in all groups. Collecting your data and presenting it to a statistician will be a good decision.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA