[EMAIL PROTECTED] (F. Goldhammer) wrote in message news:<[EMAIL PROTECTED]>... > Yes, I am expecting interesting changes of the concentration > performance in the 18 minutes, which describe a negative linear trend. > And there are differences between persons (slope). > Please could you tell me a ‘test for trends between persons‘ to get a > measure of reliability. > Thanks > Frank Goldhammer
Frank: it looks like you don't care too much for my idea about cronbach's alpha, so, let's try something else. do you have access to BJ Winer's text "Statistical Principles in Experimental Design, 2nd edition? sections 4.5 and 4.6 might be helpful. the equation on page 290, noted as r_sub_4 is called Hoyt's reliability and is an application of analysis of variance to reliability estimation. it would seem to make sense to partition the sums of squares for your 18 observations into components for trend, since the order of the observations is important. this would not, it seems to me, be the case with an ordinary set of test items where the order is usually not critical. it sounds as if you are expecting performance to deteriorate over time. so from an internal consistency standpoint, your measures might still be "reliable." it would be similar to a set of test items where the more difficult items are toward the end of the examination. as you may know, many standardized instruments are actually constructed this way - placing the easier items at the beginning and the most difficult at the end. JJ Diamond . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
