Thank you very much for you reply. I wish to compare the means of my data with the norms established by the test. The particular memory test I am using is the Wechsler Memory Scale-III which was co-normed with the WAIS- III. Each scale of the WMS-III (General Memory, Recognition Memory Visual Memory) has a mean of 100 and a SD of 15. I would like to compare the results from my data set (actually means from each scale) against those norms. I had assumed that a One-Sample T would be appropriate. To do a Z I was thinking I need the N of the normative sample (A huge stratified sample based on US census data). However my N is 50 and differences in N would be huge. Uncomfortable with that I felt that a one sample T sould be more appropriate since the normative data set against which it is being compared is so large and appears to represent the general population so well. As I look at the results printing out on SPSS, I am seeing seeing significant differences but am also wondering if there is a flaw in my statistical logic.
[posted and mailed] [EMAIL PROTECTED] (Jay Warner) wrote in [EMAIL PROTECTED]:">news:[EMAIL PROTECTED]: > May I ask, how you _know_ that the stdev = 15? Did you set it this > way? Do you have lots of data to show it so? > > I'd like to suggest that you can only _estimate_ stdev from "external" > or "internal" data. the former is when it comes from elsewhere, the > latter when it comes from the data under analysis. In the former case > we have no way of establishing the confidence of the estimate, so we > take it as "known" and use a z test. > > However you 'normalize' your data, it looks to me that you are > estimating the stdev from the data under review - thus, an internal > estimate. > > If so, then you will wind up with a t test. > > If you choose to compare a data sample against a 'norm' then you have > a one-sample test. > > Thus, a one sample t test. QED. > > If you normalize your data to reach a mean of 100 and a stdev of 15, > are you not 'giving away' some information? I believe a recent post > here advised against excessive 'normalizing' and other manipulations > of the data not related to the nature of what was being measured - the > technology. Whether or not your situation qualifies as 'excessive' > only a knowledge expert can tell, I believe. > > Cheers, > Jay > > Christopher Larsen wrote: > >> I am looking at doing an analysis of some data on memory performance >> on a standardized test of recall memory. The norms on each scale of >> the test are 100 with a standard deviation of 15. I am wanting to >> compare my data set to the normative data. Which would be a superior >> test to use, a Z test or a One Sample T-Test??? >> . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
