1
$\begingroup$

I read some studies in Software defect predictions (published in good journals) that mentioned that we should use datasets with an Events Per Variable (EPV) greater than 10; otherwise the results will be susceptible. My question is, what is the meaning of authors by the word "susceptible"? In other words, what are the disadvantages of using datasets having EPV less than 10? Do they mean, datasets with EPV > 10 provides more accurate prediction than the datasets with EPV < 10 ?

The original text of the paper I am talking about is in the following quotes: "" Larger EPV values indicate the lower risk of producing unstable results due to overfitting. Our recent work shows that defect prediction models that are trained using datasets with low EPV values are especially susceptible to unstable results""

Thanks

Improve this question
$\endgroup$
4
  • $\begingroup$ "Susceptible" has no standard statistical meaning. I suspect a full quotation from one or more of those studies, in context, might reveal what they meant by this word. $\endgroup$
    – whuber ♦
    Commented Mar 11, 2020 at 18:05
  • $\begingroup$ @whuber, yes you are right. This word might have multi-dimensional menanings.. The two sentences I am copying here from that paper are: " Larger EPV values indicate the lower risk of producing unstable results due to overfitting. Our recent work shows that defect prediction models that are trained using datasets with low EPV values are especially susceptible to unstable results" $\endgroup$
    – AAA
    Commented Mar 11, 2020 at 18:18
  • $\begingroup$ Thank you. Please include that quotation in your question. Many people here will be able to offer good explanations of overfitting, prediction, and instability in this context. $\endgroup$
    – whuber ♦
    Commented Mar 11, 2020 at 18:44
  • $\begingroup$ @whuber, thanks.. done it. $\endgroup$
    – AAA
    Commented Mar 11, 2020 at 18:47

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.