Skip Navigation
Technical Methods Report: Statistical Power for Regression Discontinuity Designs in Education Evaluations

NCEE 2008-4026
August 2008

Table 4.3.  Design Effects for RD Designs for a 50-50 Split of the Treatment and Control Group Samples


Score Distribution Percentage of the Score Distribution That Lies Above the
Cutoff Score Value (q) for p=0.50
0.50 .033 or 0.67 0.25 or 0.75 0.10 or 0.90
Normal 2.75 2.79 2.85 3.16
Uniform 4.00 3.70 3.40 2.83
Truncated Normal        
Values of k1, k2a        
-1.0, 1.0 3.65 3.50 3.36 2.99
-0.4, 0.4 3.94 3.65 3.40 2.86
-0.2, 0.2 3.98 3.67 3.40 2.84
0.0, 2.0 3.14 3.66 3.90 4.22
0.6, 1.4 3.73 4.01 3.93 3.44
0.8, 1.2 3.92 3.90 3.70 3.11
Bimodal: Equal
Mixture of Two
Normal Distributions
       
Value of lb        
2.0 5.37 3.50 3.03 2.65
1.4 3.75 3.33 3.11 2.93
0.8 2.94 2.94 2.97 3.14
0.4 2.77 2.81 2.87 3.16
Note: See the text for formulas and other assumptions underlying the calculations.
a The parameters k1 and k2 are the number of standard deviations from the mean of the full normal distribution that the left and right truncation points fall.
b The parameter l is the number of standard deviations to the right (left) of the mean of the overall bimodal distribution where the first (second) component normal distribution is centered. The calculations assume that each normal distribution is weighted equally to create the bimodal mixture distribution (that is, the parameter w=0.5).