Hello,
What would be the best methods for comparing three independent studies that used the same measurement scale (25 items Likert scale) and I only have the mean and standard deviation for each item? each group is in a different geographic location. I'm trying to check if there is a difference between the three groups that are statistically significant. but the problem is that I only have access to the means and standard divination without the actual responses and data sets.
Thank you,
Hello Yaman Sokienah Simply use GraphPad Prism software and just enter the statistics you have.
If you multiply the item mean by the item sample size, this will give you the sum of the item scale scores, (SISS). For each study, add up all 25 SISS values. This will give you the total scale score (TSS) for each of the three studies. Divide the three separate TSS values by the respective sample sizes to determine the mean TSS for each study. It is these mean TSS values that you can compare by inspection against each other. In order to do so using ANOVA, you will need the three standard deviations, which doesn't appear that you have.
Are the sample sizes large, or are you willing to assume normality?
I assume these are old studies so that the authors are no longer alive to email.
There is no access to the raw data. The sample soze is ranging between 75-150.
There is no access to the raw data. The sample soze is ranging between 75-150.
With the sample size you can calculate the (1-p)x100% confidence interval for of the means in each one of the studies. If p=0.05 you will have the interval with 95% probability for each study results. If interval confidence overlap there is not significant difference; otherwise, the results will be statistically different.
I would try simulation. You can generate many samples of 70 to 150 from a distribution with a mean and standard deviation listed. Additional complexity could be added by using different distributions (not just Gaussian). R is great at this sort of thing, but other choices would be equally effective. A key discussion point is how accurately you feel a sample size of 150 models the underlying distribution.
Actually means.and standard
Deviations don't tell you much about likert.variables. see the attachment to tell you
what might be used instead.. however if you don't have access to the original data I would choose a different project. Sorry I can't be more positive.. Best wishes David Booth
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS