You should apply non-parametric test when sample is less than 30

Not only that, some time simulation also needed to verify your results.

You can go for t-test although you have to check the assumptions if they are all ok.

But if your data do not follow the assumptions it is better to go for non-parametric tests like should be the choice.

the article may help you

http://dx.doi.org/10.1016/j.jeconom.2015.03.004

The appropriate for your case is non-parametric test because sample is less than 30.

Good luck!

You should apply non-parametric test because the sample is less than 30.

The test

When sample is less than 30 , should do T.Test by hand or Excel or SPSS programs to test Your hypotheses 

You can see this video for more details, it's a great explanation ( math guy zero )

https://youtu.be/URF-Cw5Bx6c

I endorse the comments the other expert and maybe you'll get into trouble for normality. So use histogram grafic

I would suggest to use a scale between 10 (high) and 1 (low) for every item. Now you can compare them by median and average standard deviation. You of course have to operationalize a bit but if you focus on around 20 correlation factors/items you'd having a valid analysis.

We do the same with our small panels >10 with great success.

You can go for non parametric test because your sample size is less than 30.

Salam Bosy. It is difficult to recommend specific statistical analysis with the absence of any clue on the nature of your investigation. However, in general terms, the panel data analysis you would like to use should have both cross-sectional dimension (and you have this one as you mentioned, 18 companies) and time-series dimension (for a certain number of years, for example). By then, you will have a panel data set. How many variables you have will have also an influence on the appropriate sample size. Apparently, having the data of eighteen companies may be sufficient, depending on what you are going to find out, again. Therefore, I recommend you either collect more data on these companies to have a panel data (so can use panel data regression models) or to use the data you have currently collected, which seems to be cross-sectional provided that it is appropriate and sufficient to provide evidence for your arguments. 

Hope that helps 

Hi Bosy!

Are you analysing IPO returns as dependent variables with various factors as independent or control variables? If that is the case you can perform panel regression analysis even with data of 18 companies provided you have sufficient number of factors as explaining variables. More clarification is possible only after knowing the exact nature of your investigation or by looking at your dataset. Hope this answer helps you.