راهکارهای درمانی خاص در درمان لکنت زبان کلینیک تخصصی لکنت کرج کلاک بالا – خیابان نیلوفر ۳ شرقی
when calculating either rate or percentage of stutter¬ing. Acceptability of error between two counts of the same data can be calculated using the formula given in Figure 9.3. Generally, for both inter¬rater and intra-rater measures, the accepted minimum reliability figure is 90 percent (Boberg & Kully, 1985; Prins & Hubbard, 1990). Pearson product- moment correlations may also be applied to test inter-rater agreement. Druce, Debney, and Byrt (1997) suggest correlations above. 0.8 as being excel¬lent, whilst Ryan and Van Kirk (1995) cite. 0.9. To return to our example, then, 30/33 x 100 = 90.9 percent agreement, which is satisfactory (although higher figures may be required for some research purposes). As we see from Figure 9.3 our independent raters here do not meet the 90 percent criterion, and the passage would need to be re-analyzed until a higher level of agree¬ment is reached.
Kully and Boberg (1988) have pointed out that inter-rater, and particu¬larly inter-clinic reliability is often poor. Nonetheless, highly trained judges can achieve consistently high levels of agreement between observers (e.g., O’Brian, Packman, Onslow, & O’Brian, 2004; Young, 1975). O’ Brian et al. (2004) compared the reliability of the fluency count with a nine-point sever¬ity scale, finding very high levels of intra-judge and inter-judge agreement for both measures. Clinicians need to be aware that although two counts of the same data set may yield similar or even identical intra-/inter-rater figures, these do not necessarily represent the identification of the same moments of stuttering. So, theoretically, two raters might identify an identi¬cal number of disfluencies but with completely different loci, and this dis¬crepancy would not be revealed by the reliability count. Event by event consistency is a significant difficulty and probably the best course of action is to have all inexperienced raters undergo repeated practice with highly
Procedure: Collect the two calculations of the data. Divide the smaller of the two numbers by bigger number and then multiply by 100 to find the percentage.
Worked example: Rater 1 calculates a total of 175 syllables spoken, rater 2 calculates 200