Pingu wrote: > > Hello all statistics altruists - > > I have run a between subjects experiment testing subjects' speed of > dialling from a mobile phone under 2 different conditions. 12 subjects > each dialled the same 10 names under both conditions and i want to do > run t-tests in SPSS to see if there is an effect of condition on > dialling time. > > I am pretty sure that i need to be running a 2-tailed "paired samples" > t-test in SPSS to do the analysis i want. The problem is i am not sure > that i have set out my data correctly for this. > > At the moment i have 2 columns representing the 2 conditions. Each > column contains 120 rows of data (12 subjects x 10 dial attempts). I > then choose [column 1 vs column 2] in the "paired samples t test" box > of SPSS. > > Is this correct? I have the feeling that this set out may be treating > the data as if there were 120 participants instead of 12. Im not even > sure if this makes any difference, but i am a novice with very little
That's not correct. You can't treat the trials/names as independent. Either: a) use two columns of 12 data points (1 per subject) made up of the mean or median for the 10 names (use the paired t command) b) use repeated measures ANOVA and have 12 rows and 20 columns. Use the GLM repeated measures command and define two within subject factors (one with 2 levels and one with 10 levels). Data from the same person is kept on the same row/case. Strictly, the b) is better as it use more of the raw data (though one I argue that the test is generous since it treats the "names" as fixed factors). Thom . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
