I've done an experiment with human participants, n=40, so it's a large sample n>=30 from a large population. Now I'm trying to understand the statistical significance of my results and verify that I am interpreting things properly. This type of study is NOT my area of expertise.
I'm basically looking for clarification on the following points, for somebody to read what I've done and give me a reason why it's OK or why it's not OK. I appreciate the help so much.
First off, if I've asked participants which of A,B or C do they "most prefer", and I have gotten results back along the lines of 68%, 20%, 12% respectively. I can then say pretty easily that "with a 98% confidence level, A will be the most preferred for the majority of the population", using z scores. This I am confident about. But can I say, with some even greater level of confidence, that A will be most preferred by more of the population than either B or C (but not necessarily the majority of the population)? A plurality of the population will prefer A, basically. Is that possible, and how would I do that?
Now let's say I've asked participants to rank preferences with numerical values, so I've had 40 people give me there "rankings" from 0-7 for X and Y. And so I can have an average an standard deviation for X and Y values. And I can construct confidence intervals, where we can say with 95% confidence that X will be within some interval around X, and same idea with Y.
So, for instance:
X = 80 +/- 10
Y = 50 +/- 15
with 95% confidence level.
Then, can I say that because 50+15=65 and 80-10=70, that I have 95% confidence that X>Y? I want to say this, but I think I can't.
When I google comparing means of samples, they talk about having two different populations, and they usually talk about sampling the same variable from two different populations. What I want to do is compare two different variables, but taken from the same sample. How do I do that?
My only idea is to subtract pairwise the value of every X,Y obtained from the participants. So if my data was:
Particpants 1 2 3 4
X = 10 11 12 14
Y = 11 10 14 12
X-Y= -1 1 -2 2
I could then take the average and standard deviation of X-Y, and if the confidence interval I constructed did not contain 0, then I could say X>Y with x degree of confidence. Is this the case?
Thanks so, so much to anyone that can point me in the right direction here!