## ------------------------------------

## Updated summary statistics can be found here: https://doi.org/10.13020/przg-dp88 from Saunders, G.R.B., Wang, X., Chen, F. et al. Genetic diversity fuels gene discovery for tobacco and alcohol use. Nature 612, 720–724 (2022). https://doi.org/10.1038/s41586-022-05477-4

## ------------------------------------
Summary statistics based on Liu, M., Jiang, Y., Wedow, R., Li, Y., Brazel, D. M., Chen, F., ...., Vrieze, S. (2019). Association studies of up to 1.2 million individuals yield new insights into the genetic etiology of tobacco and alcohol use. Nature Genetics, 237-244 (2019). https://doi.org/10.1038/s41588-018-0307-5

Full meta-analysis of all GSCAN cohorts summary statistics except 23andMe summary statistics. These set of results have been pruned for sites with minor allele frequency greater than 0.001, have an effective sample size of at least 10% of the maximum sample size, overall imputation quality (Effective_N/N) > 0.3 and  present in 20 or more studies. The results are also annotated using TabAnno (https://genome.sph.umich.edu/wiki/TabAnno) with UCSC genome browser annotations. We did not include allele frequency for data privacy reasons and instead used HRC allele frequency and 1000G allele frequency where HRC allele frequency is not available for the variant.
 
Header Information:
CHROM: Chromosome
POS: Position
RSID: rsID
REF: Reference allele
ALT: Alternate allele
AF: Allele Frequency from HRC or 1000G
STAT: Chi-square statistic
PValue: p-value
Beta: Beta based on the alternate allele
SE: Standard error of the beta
N: Sum of sample size across contributing cohorts
Effective_N: Sum of Sample size * imputation r2 across contributing cohorts
Number_of_Studies: Number of Studies at this site
ANNO: ANNO tag displayed the most important variant type that happened at GENE. Please see https://genome.sph.umich.edu/wiki/TabAnno for more information.
ANNOFULL: ANNOFULL tag is the full set of annotation. Please see https://genome.sph.umich.edu/wiki/TabAnno for more information.

##Update on September 23rd 2019
Previous betas are reported based on the Wald statistic. Given the different ways of how each trait has been measured (binned, normalized, etc), we have since changed beta to be based on the chi square statistic instead using  the following transformation (below "beta", "statistic", and "af" refer to the respective column label in the summary statistic files):

For continuous traits with assumed standard deviation of 1 for the phenotype:
beta.est <- sign(beta.ori)*sqrt(statistic)*sqrt(1)/sqrt(2*N*af*(1-af));
beta.sd <- sqrt(1/(2*N*af*(1-af)))

For binary traits (var.y is the prevalence of cases. For our analysis, we took the mean prevalence across all studies in the analysis):
beta.est <- sign(beta.ori)*sqrt(statistic)/sqrt(2*N*af*(1-af))/sqrt(var.y));
beta.sd <- sqrt(1/sqrt(2*N*af*(1-af)*var.y));