query re normalising data for an unpaired (indep. samples) t-test Watch
this probably could be a sixth form level question but unsure...
I have a some data for a variable I'd like to compare. Prior to running the t-test in SPSS, I split the variable so I could see histograms of the variable in each category.
To keep it simple, let's say I'm comparing height in boys and girls. I split height by boys and girls to gain a histogram for each. The histogram for boys is pretty much normally distributed whereas the histogram for girls is negatively skewed.
I take it when normalising data (in this case I'd be doing an x^2, x^3 or e^x transformation to try and normalise a negative skew) I would need to apply the same transformation across the entire variable not to just the boys/girls? Or is it possible to normalise only one part of the data?
If do this and it as results in the other sex's distribution being skewed, would it mean I can't I do a t-test and should do a Mann-Whitney U test instead?
Also on a separate note: I've collected data where I've picked up measurements over time. I've then taken the median and inter-quartile range. When comparing the inter-quartile range values (these are my summary values for the dataset) between the sexes, would I carry out a Mann Whitney U test by default seeing as the data they originated from are not normally distributed? Or would I check (and try to correct for) a normal distribution for the range of IQR values and perform an unpaired t-test?
While you're waiting for an answer, did you know we have 300,000 study resources that could answer your question in TSR's Learn together section?
We have everything from Teacher Marked Essays to Mindmaps and Quizzes to help you with your work. Take a look around.
If you're stuck on how to get started, try creating some resources. It's free to do and can help breakdown tough topics into manageable chunks. Get creating now.
Not sure what all of this is about? Head here to find out more.