For each outcome, results are reported separately for males and females for each of the
three levels of schooling (JSS 1 through JSS 3). Effects attributable to the programmes are estimated by the two interaction terms. The interaction between wave and FLHE provides an estimate of how much students in the FLHE sites changed between waves, compared to students in the delay sites. The interaction between wave and FLHE+C estimates how much students in sites with both FLHE and the community programme
changed, again compared to those in delay sites. It is the interaction terms that are of primary interest, since they answer the question of how much the FLHE and the FLHE+C programmes achieved. Given that there are 12 tests of programme effects
for each outcome (2 programme effects for each of males and females in each of three grades), if results were strictly random, one or more effects significant at .05 would be expected through random fluctuation 46% of the time. Taking this into consideration, we pay attention not only to statistically significant results, but give greater weight to results significant at .05 or lower at one school level that are found together with results in
the same direction at other levels of schooling. While the probability of getting one or more significant results at three different school levels is 143, the probability of getting one or more significant results along with two others in the same direction is only .0375, raising confidence that results were not due to random fluctuation. This strategy has been used elsewhere in evaluation studies involving multiple outcomes to guard against reporting results which have a high probability of being produced by random fluctuation.