Yes, you have to present p value and CI interval.

I assumed you are used paired t-test to examine the intervention effect. My understanding is that the associated p-value indicates the possibility of no intervention effect based on your data set. It is always useful to also construct confidence limits for the true mean difference.

Yes, Exactly you need to to present p value in order to know the significant difference and CI interval.

I prefer the CI, because it provides the magnitude of the data

I did use an unpaired t-test. My data is

Group 1: n=44 , mean value = 65.39, SD 7.38

Group 2: n=46, mean value = 59.70, SD= 13.1, p=0.0134

From this, how do I calculate the significant difference and CI interval?

Based on what you described in your first post, you had a design that should have produced paired (or matched) data. Did you have two measurements from each student - one score before and one after inrtervention? If you did then you have paired data. If you have paired data, you should carry out a paired t-test and obtain a single confidence interval for the mean of the differences. Hope this helps.

What kind of distribution shows your data? parametric or not parametric distribution? What kind of design are you using? I think that you have to show your CI because the audience can see the significance of your data, you can use t-test or u-test, you can do so many things with your data!... do you use the same students for both measurements? because you have different "n" in both groups, what happened?

yes you can appliey cI

How did you measure anxiety and memory? That determines how you analyze your data. I agree with one of the comments, yours is a before and after intervention study and must be analyzed by a before and after analysis technique (paired t-test for continuous outcome and McNemar’s chi test for within subjects design for binary outcome-remember there are assumptions to be checked. Definitely we need to see the confidence interval due to its ability to show precision and it communicates more than the traditional p-value. Hope this helps

Be aware that small differences will have significant p-values if the compared groups are very large , adding the the confidence will allow to see the flluctation of your proportion or ratio, large confidence intervals often do not allow adequate clinical use. If you want to work with critical values or cutt off levels then you must look for methods using receiver operating characteristic curves and work with sensitivity and specificity based on fals and true psotives and negatives.

For Jason, to calculate your intervals go to http://www.mccallum-layton.co.uk/stats/ConfidenceIntervalCalcProportions.aspx

Though you need to find the significance difference between groups you can also find the C.I with p-value .It gives you how much significance mean diff. among groups before and after

Thanks everyone. The groups were unpaired, but I have been able to calculate a CI level from the mean, n, and p-value. It was extra work, but now the data is that much more descriptive.

Based on your description, the only way I can think of where your data would be unpaired is if you did not have any identifier on the 90 subjects. If your data were really unpaired, then what Andrew said above is right, as you will not be able to distinguish the treatment effect from the group effect.

I prefrer CI which may give you the magnitude of the uncertainty.

They are unpaired in that each group was measured once. 46 were in the treatment group and 44 were in the control group. The paired test is for testing before and after in the same subjects (as I understand from http://www.gpnotebook.co.uk/simplepage.cfm?ID=x20050323103508411760). Each group was independent from the other. Not all research is done with the same group before and after (paired). That's why a sufficient N can reduce the group difference effect. I thought that was the whole point of statistical analysis, because in any evaluation of 2 different groups, the differences found could be due to chance. The higher the N, the less likelihood that the difference is due to chance alone and the greater likelihood that the difference was due to the intervention. I thought that was the whole point of p-values and CI levels. Is that not right?

Dear Jasan bussell. I request you to try with correlation and regression analysis for the frequencies you have selected in the study.

You may use both but the CI is more informative. If i was to choose between two i would go for CI