Validity Tips
face validity (logical validity)--administer to groups you have good reason to believe will answer in an expected way--see if they do
uni-dimensionality (all items measure same thing/construct?
amount of variance (what items divide respondents?)
look at bi-viariate relationships among items to see if they are related empirically to each other--if perfectly related, why have both items?
examine multivariate relationships to eliminate “deadwood” items--adds to power of scale. Does each item represent a slightly different variation of the same variable?