The survey of Lyman \alpha emitters ( LAEs ) with Subaru Hyper Suprime-Cam , called SILVERRUSH ( Ouchi et al .
) , is producing massive data of LAEs at z \gtrsim 6 .
Here we present LAE simulations to compare the SILVERRUSH data .
In 162 ^ { 3 } comoving Mpc ^ { 3 } boxes , where numerical radiative transfer calculations of reionization were performed , LAEs have been modeled with physically motivated analytic recipes as a function of halo mass .
We have examined 2 ^ { 3 } models depending on the presence or absence of dispersion of halo Ly \alpha emissivity , dispersion of the halo Ly \alpha optical depth , \tau _ { \alpha } , and halo mass dependence of \tau _ { \alpha } .
The unique free parameter in our model , a pivot value of \tau _ { \alpha } , is calibrated so as to reproduce the z = 5.7 Ly \alpha luminosity function ( LF ) of SILVERRUSH .
We compare our model predictions with Ly \alpha LFs at z = 6.6 and 7.3 , LAE angular auto-correlation functions ( ACFs ) at z = 5.7 and 6.6 , and LAE fractions in Lyman break galaxies at 5 < z < 7 .
The Ly \alpha LFs and ACFs are reproduced by multiple models , but the LAE fraction turns out to be the most critical test .
The dispersion of \tau _ { \alpha } and the halo mass dependence of \tau _ { \alpha } are essential to explain all observations reasonably .
Therefore , a simple model of one-to-one correspondence between halo mass and Ly \alpha luminosity with a constant Ly \alpha escape fraction has been ruled out .
Based on our best model , we present a formula to estimate the intergalactic neutral hydrogen fraction , x _ { HI } , from the observed Ly \alpha luminosity density at z \gtrsim 6 .
We finally obtain x _ { HI } = 0.5 _ { -0.3 } ^ { +0.1 } as a volume-average at z = 7.3 .