it doesn't matter what you do, but it does matter who does it!

Post on 23-Feb-2016

56 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

DESCRIPTION

It doesn't matter what you do, but it does matter who does it!. Martin Shepperd, Brunel U Tracy Hall, Brunel U David Bowes, U. of Hertfordshire. Overview. Many empirical studies (200+) to predict software faults No technique dominates - PowerPoint PPT Presentation

TRANSCRIPT

1

It doesn't matter what you do, but it does matter who does it!

Martin Shepperd, Brunel UTracy Hall, Brunel U

David Bowes, U. of Hertfordshire

2

Overview

• Many empirical studies (200+) to predict software faults

• No technique dominates• Conduct a meta-analysis to explain variation in

the results• Used factors of (i) classifier (ii) metric type (iii)

data set (iv) research group

3

Systematic Review

• Conducted by Tracy Hall and David Bowes– T. Hall, S. Beecham, D. Bowes, D. Gray, and S. Counsell. “A systematic

literature review on fault prediction performance in software engineering”, Accepted for publication in TSE (download from BURA).

• Located 208 relevant primary studies• Due to reporting requirements used 18

studies that contain 194 results– binary classifiers, confusion matrix, context details

4

(i) Classifier

5

(ii) Metric Type

• Delta• Static• Process• Other• Combinations

6

(iii) Data Set

ECLIP :93EMTEL :26MOZ :25COS :16EXCL : 9VISTA : 4(Other):21

7

(iv) Research Group

8

Response variable

Response variable – Matthews correlation coefficient (MCC)• stable (uses all 4 cells of the confusion matrix)• easy to interpret (0=random)• easy to compare• related to chi-squared test

9

Matthews correlation coefficient

10

ANOVA model4-way linear random effects model with

interactions

11

ANOVA Results

Factor % of varAuthor group 61%Metric family 3%Author/metric 9%Everything else 8% (but not significant)Residuals 19%

12

Confounders?• autocorrelation• system age• pre /post-release data collection

not significant• Homogeneity of variances

robust Levene test p=0.51• Normality of the RV

slight +ve skew (0.12) and leptokurtosis (0.26)

13

Conclusions

• There are problems with how research is replicated– expertise– bias

• Search to– de-skill– de-bias

14

Final word

We cannot ignore the fact that the main determinant of a validation study result is which research group undertakes it.

top related