While it is possible to use a one-way ANOVA to determine whether there is a significant difference, I would recommend using a paired sample t-test for a post-pre comparison to match respondents if possible.

Regardless, the choice of statistical method--whether one-way ANOVA or paired sample t-test--should have no bearing on the scoring. If you were not expecting a negative difference (scores worsened afterwards), then inspect the survey procedure or the process being evaluated to determine reasons for this unexpected result.

Yes, you can. You can refer to the textbook " Design and Analysis of Experiments" Douglas Montgomery - chapter 5 / Example 5.3.

Best of luck

Thanks Emmanuel,

I have already done paired t test for significance.

My question is can I Use ANOVA for negative values? 

Thanks Gerges for your response.

Chapter 5 Example 5.3  of  is of  "Design and Analysis of Experiments" Douglas Montgomery is for Two Factor design.

I could not relate it to my statement.

Don't use ANOVA on survey data--not valid.

Hello Bharat,

Do you mean that your scoring system has negative values? If so, ANOVA may not be ideal since it's usually meant for ratio-level data (with an absolute zero)--see the link below.

I am puzzled though as to why you would have negative test scores unless you penalize students for wrong answers, and they have a lot of wrong answers. My advice would be to (a) use paired samples t-tests instead of ANOVA and (b) use a scoring system whose lowest possible value is zero.

http://www.statsoft.com/Support/Forum/forumid/3/postid/2023/scope/posts