SED 374 Chapter : September 15th, 2014
Document Summary
Reliability ( r ) = extent to which the test is stable or consistent, reliable, to be valid, it must be reliable. Higher r = greater confidence test is accurate correlation coefficients range from -1 to +1 neg. inverse relationships: as one score goes up, the other goes down pos equal relationships: as one score goes up, so does the other. When it equals 0 there is no consistency we look for +8 to +95 range. Higher stakes- we want a higher r score. Test-retest - same test given to same group twice, compare results. Confidence interval: range of scores where a child"s true score would fall if there were no error in admin, there is always error though. Inter-rater reliability - usually when using observation (behavior) 2 or more assessor, compare ratings, want high reliability. *for something to be valid it needs to be reliable*