T-test or Pearson's r?

[joe7987]

Hello all,

I'm a graduate student, finding my way through research statistics. I just completed data collection of an experiment that examined human assessment and automated assessment of the same material.

My goal, now, is to see how similar human assessment and automated assessment were on this task. I already conducted a t-test and found that there was no significant difference. Would it be correct to then calculate a pearson's r to see how similar they are? I know that a correlation tells you the relationship between two variables (if one changes, what happens to the other.. positive, one goes up (or down), the other goes up (or down).. negative, one goes up (or down), the other goes down (or up), etc...)

Does this fit my purpose? I have seen correlations used for purposes of looking at two different events (SAT scores compared to college performance), but never looking at two different ways of doing the same event.

Sorry for the drawn out post, I just wanted to make sure I was clearly explaining myself. Thanks in advance!

[Dason]

Did you take into account the pairing when you did the t-test?

[joe7987]

Yes. It was a paired t-test.

[Karabiner]

My goal, now, is to see how similar human assessment and automated assessment were on this task. I already conducted a t-test and found that there was no significant difference.

This statement is quite uninformative without reporting the actual p-value
and of course your sample size.

Would it be correct to then calculate a pearson's r to see how similar they are?

It depends what you want to understand by "similar". For example

1 - 11

2 - 12

3 - 13

4 - 14

are pairs of values which show a perfect correlation, but
they are not similar (show large differences).

Kind regards

K.

[joe7987]

Sample size is 57. There was no significant difference at alpha = .05.

I mean similar on a pair by pair basis. For instance.. if I have two methodologies (A and B) of performing task Z. I perform task Z with methodology A 57 times. I then perform task Z with methodology B 57 times. Do A and B produce the same performance on task Z?