> source("fsr.fast.txt")
> fsr.fast(x=ncaa2[,1:19],y=ncaa2[,20])->out
   var   pval  pvmax    Rsq  ghigh   glow
1    2 0.0000 0.0000 0.7069 0.0000 0.0000
2    3 0.0001 0.0001 0.7539 0.0006 0.0004
3    5 0.0116 0.0116 0.7708 0.0656 0.0270
4    4 0.0053 0.0116 0.7901 0.0270 0.0270
5    7 0.0025 0.0116 0.8110 0.0270 0.0270  <-- jumps through .05, 5 var. model chosen
6   17 0.0433 0.0433 0.8197 0.1011 0.0804  <-- for gamma0=.05
7   15 0.0527 0.0527 0.8274 0.0979 0.0791
8    6 0.1056 0.1056 0.8327 0.1584 0.0864
9    9 0.0826 0.1056 0.8386 0.0864 0.0864
10   8 0.0536 0.1056 0.8457 0.0864 0.0864
11  12 0.2350 0.2350 0.8484 0.1922 0.1566
12  10 0.2864 0.2864 0.8505 0.1910 0.1542
13  13 0.3163 0.3163 0.8524 0.1703 0.1054
14  18 0.2697 0.3163 0.8546 0.1054 0.1054
15  11 0.4953 0.4953 0.8555 0.1651 0.1238
16   1 0.6326 0.6326 0.8559 0.1582 0.1116
17  14 0.7056 0.7056 0.8562 0.1245 0.0784
18  19 0.8605 0.8605 0.8563 0.0956 0.0453
19  16 0.9032 0.9032 0.8563 0.0475 0.0000
> out
$mod

Call:
lm(formula = y ~ x)

Coefficients:
(Intercept)          xx2          xx3          xx5          xx4          xx7  
   -42.1069       3.4714       0.2391       0.2787       0.6770      -2.5913  


$size
[1] 5

$x.ind
[1] 2 3 5 4 7

$alphahat.ER
[1] 0.02142857   Compare to alphahat.ER=.044 to .046 chosen by regular FSR
                 with phony variable generation.