Question 514:
1Answer:
No answer provided yet.From what you've indicated here, your approach seems very reasonable. You've said you've matched the pairs of respondents for before and after the event and have results before and after from a survey. When you conduct a paired-t test, you are essentially removing the between subject variability and are instead looking at the difference between scores (called a within-subjects design). You could in-fact get the same results by subtracting score 1 from score 2 and conduct a 1-sample t-test on this difference.
Now, if you conduct the paired t-test, you will have an average before and after. The t-test will be on the difference between the two and you will get a p-value. From this p-value it is definitely appropriate to make the judgment on whether there was a difference before and after the event. I'm not sure what the event is, but when making the conclusion that the event had a positive impact, be sure you're accounting for other nuisance variables and spurious correlations such as another salient event the subjects might have encountered or different days of the week, year, temperature changes etc.
Regarding 1 or 2-tailed, strictly speaking, this is something you're generally supposed to decide prior to the test being conducted. In practice, it is best to use the 2-tailed paired t-test. Only use the 1-tailed test if you have strong evidence prior to the testing that you expect the results to increase (or decrease). With your sample size above 60 using the within-subjects setup, you'll likely have a reasonably good chance of detecting a difference (higher power). In fact, this is one of the main advantages of the within subjects approach (paired-t-test)--you get more power with a smaller sample size.
If you'd like me to help analyze your data or you'd need more assistance with the computations or interpretations I'm happy to help as I have a lot of experience analyzing survey data. You can send me the data-file in an anonymous format if needed.