Empirical Bachelor Thesis - Normality Distribution or Not? Allowed Methods - Please Help

katxt

Well-Known Member
#2
It looks like you have 14 data points for each metric. It is almost impossible to determine if data is normal or not from so few points. A "not significant" result of a test is simply saying "I don't know if it is normal or not."
Skewness is of limited use, though it does fill up reports. Boxplots are good for odd data. I would use non parametric correlation, and quote the standard error instead of a confidence interval.
 

AngleWyrm

Active Member
#3
Here's an illustration of a complete network graph composed of 14 nodes
graph.png
Is this what you're referring to, or do you mean 14 separate network graphs composed of an unspecified set of nodes and edges?
 

katxt

Well-Known Member
#5
If you look for significant correlations, be careful. With 10 metrics you have 45 possible p values. You are likely to get false positives all over the place. Distinguishing between true and false positives has always been a thorny problem which hasn't really been solved.