----------------------------------------------------------------------------------------
META-ANALYSIS SUMMARY STATISTICS
----------------------------------------------------------------------------------------
chr 		- this is the chromosome
snp			- the marker name (rsID)
bp			- the basepair position (hg19)
a1 			- the first allele for this marker in the first file where it occurs
a2 			- the second allele for this marker in the first file where it occurs
freq 		- weighted average of frequency for a1 across all studies 
b 			- overall estimated effect size for a1 
se 			- overall standard error for effect size estimate 
p 			- meta-analysis p-value
direction	- summary of effect direction for each study, with one '+' or '-' per study 
het_I2 		- I^2 statistic which measures heterogeneity on scale of 0-100%
het_Chi2 	- chi-squared statistic in simple test of heterogeneity
het_p 		- P-value for heterogeneity statistic
cases 		- number of cases in which the marker is succesfully genotyped/imputed
controls	- number of controls in which the marker is succesfully genotyped/imputed

----------------------------------------------------------------------------------------
LMM SUMMARY STATISTICS
----------------------------------------------------------------------------------------
chr 		- this is the chromosome
snp			- the marker name (rsID)
bp			- the basepair position (hg19)
a1 			- the first allele for this marker in the first file where it occurs
a2 			- the second allele for this marker in the first file where it occurs
freq 		- frequency for a1
b 			- estimated effect size for a1 on LMM scale, for conversion to OR see below
se 			- overall standard error for effect size estimate  on LMM scale
p 			- LMM p-value

----------------------------------------------------------------------------------------
SCALE OF LMM EFFECT SIZE (b)
----------------------------------------------------------------------------------------
The effect size estimate from the LMM is not a "regular" beta. This means that converting
it to an OR is NOT simply OR = exp(b)! To convert the LMM beta (b_lmm) to an OR use
this approximation:

Ncases      = 12577
Ncontrols   = 23475
Pcase       = Ncase / (Ncase + Ncontrols)
OR          = ((Pcase + b_lmm)/(1 - Pcase -  b_lmm)) / (Pcase/(1 - Pcase))

The "regular" beta is than just ln(OR).
The same approximation works to convert the standard error (se).

----------------------------------------------------------------------------------------
RELATION TO TABLE 1 OF THE MANUSCRIPT
----------------------------------------------------------------------------------------
Table 1 of the manuscript summarizes the association statistics for all new genome-wide
significant hits in the discovery phase.
The OR in the discovery phase is that of the LMM analysis. So it was calculated using
the approximation above.
The OR in the replication phase is from the meta-analysis in our replication cohort.