On 6 Mar 2002 06:17:09 -0800, [EMAIL PROTECTED] (Mobile Survey) wrote: > Dear all, > I am using the following 7-point semantic differential scale scale > (with mid-point anchor too) in my survey. > 1. extremely dissatisfied > 4. neither dissatisfied /nor satisfied > 7. extremely satisfied > I understand that this can be treated as a metric interval scale. The > problem however, is that the scatter plots of this scale with another > 7-point scale on another variable that I am using obviously does not > look like the continuous data that we need to have a good linear > relationship where the Pearson correlation coefficient makes sense. I > had the following question
I think -- it is 'semantic differential' when you have several topics, each of which is subjected to ratings on 10 or 20 bipolar pairs (10 or 20 'items'). Try one I just found, http://faculty.ncwc.edu/toconnor/308/308lect05.htm If one item has a terrible distribution, it is going to correlate poorly. Okay, that means, Drop it. For the statistics, you drop both: bad content and bad wording. (Wording is usually easier to mend.) If it has a terrible distribution, it does not qualify as a wholly suitable Likert item, either. > 1.In this case should I use the Spearman rho coefficient? > 2.If the above scales are for indicators of a latent variable , then > should I report the inter-item and item to total correlations and > their significance in terms of Spearman correlations? No. No. -- Rich Ulrich, [EMAIL PROTECTED] http://www.pitt.edu/~wpilib/index.html . . . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . ================================================================= . .
