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

NCEE 2008-4026
August 2008

Table 7.2.  Required School Sample Sizes to Detect Target Effect Sizes, for Various RD Designs


Assumes a Two-Tailed Test, a Value of .15 for the Intraclass Correlations, the Cutoff Score Is at a
Tertile of the Normal Score Distribution, and a 2:1 Split of the Research Groups
Unit of Treatment Assignment: Other
Fixed or Random Effects
Number of Schools Required to Detect an
Impact in Standard Deviation Units of:
.20 .25 0.33
I: Students Within Sites (Schools or Districts): Site Effects Fixed
R2 = 0 39 25 14
R2 = .2 31 20 12
R2 = .5 20 13 7
R2 = .7 12 8 4
II: Classrooms Within Schools: Fixed School Effects
R2 = 0 142 91 52
R2 = .2 114 73 42
R2 = .5 71 45 26
R2 = .7 43 27 16
III: Schools: Within Districts: No Random Classroom Effects
R2 = 0 359 230 132
R2 = .2 288 184 106
R2 = .5 180 115 66
R2 = .7 108 69 40
IV: Students Within Schools: Random Site Effects
R2 = 0 63 40 23
R2 = .2 50 32 18
R2 = .5 31 20 12
R2 = .7 19 12 7
V: Classrooms Within Schools: Random School Effects
R2 = 0 166 106 61
R2 = .2 133 85 49
R2 = .5 83 53 30
R2 = .7 50 32 18
VI: Schools Within Districts: Random Classroom Effects
R2 = 0 462 296 170
R2 = .2 370 237 136
R2 = .5 231 148 85
R2 = .7 139 89 51
Note: See the text for formulas and other assumptions underlying the calculations. The figures assume that the assignment scores are normally distributed.