Small Sample Size Confidence intervals difference in mean

#1
Hi everyone. I am currently working on report and have to analyse the 90% CI for differences in means. The sample size is 20 and the data is comparing weight before and after an intervention. So this is a paired example if I am correct.
I have checked and the data is approximately normally distributed. I can use the t-distribution and t test in STATA. For the small sample size am I able to use the normal STATA function ttest groupa==groupb, level(90) as per the normal ttest or do I need to use a different one due to the small sample size?
Thanks so much for the help.
AL
 

hlsmith

Less is more. Stay pure. Stay poor.
#2
Is this for an assignment? If not, I'd conduct a one-sample ttest of differences versus zero and slap the 90% CI on the mean differences.
 
#3
Is this for an assignment? If not, I'd conduct a one-sample ttest of differences versus zero and slap the 90% CI on the mean differences.
Hi hlsmith, do you mind clarifying a little. This is for a report. its hypothetical not a real life analysis. I was under the assumption the data is paired as its the same person at two different times (before and after) and therefore a paired t test at 90 level needs to be done. Not a two sample test unpaired at 90%
 

hlsmith

Less is more. Stay pure. Stay poor.
#4
I wrote one-sample not two-sample. I would report the mean differences with standard deviation value and slap 90% CI on mean estimate and show it excludes the null.

Just curious where your 90% value is coming from? Traditionally 95% used and with small sample making it say 99% would help with assuring generalizability.
 
#5
yeah i know. we usually use 95%. may just be a test to see if we know how to change it and interpret it. and it does explude 0 so that's good=)
thanks