Most notable changes per version (minor changes and bug fixes are not explicitly listed)

v1.09
- added an option for using weights with the burden scoring in gene analysis
- added an option to include or exclude individuals from the input data by individual ID
- fixed a bug that caused missing NA values for MAC in the .genes.raw files when using burden score analysis (1.09a)

v1.08b
- fixed considerable slow-down of the SNP-wise Multi model that resulted from the introduction of the revised SNP-wise Mean model in v1.08
- fixed bug in meta-analysis code where input correlation values were not squared as intended (fixed in v1.08a)
- fixed bug in raw data gene analysis with chromosome X in some circumstances when there were gender missing values present in the data (fixed in v1.08bb)

v1.08
- changed the SNP-wise Mean model (see details on MAGMA site)
  - test statistic was changed from sum of df=2 chi-squares to df=1 chi-squares across SNPs
  - evaluation was changed to numerical integration using the Imhof approach, with an adaptive permutation procedure as backstop
- added option to gene meta-analysis specify cross-cohort correlation matrix to account for sample overlap

v1.07b
- fixed segmentation fault when reading gene annotation file during gene analysis
- added additional columns to per-gene output files created during gene set/gene covariate analysis (for use with QC R scripts)
- fixed missing 'nonhuman' settings (1.07bb)

v1.07
- fixed occasional segmentation fault bug occurring on some Mac and Windows systems when using --pval
- general restructuring of code and various fixes and tweaks
- improved memory performance when using --pval option
- complete rewrite of the gene-level analysis component
- implementation of a range gene-level  (gene set, gene covariate) conditional, joint, and interaction analysis options
  - moved all existing gene-level model options to the --model flag
- added additional per-gene output from gene-level analysis to assist in post-hoc inspection of results
- deprecated and removed the gene-level FWER empirical multiple testing correction option
- added modifiable variable name abbreviation option

v1.06
- added option to read in synonymous SNP IDs from file, to map SNPs designated with different but synonymous SNP IDs in different input files
  - provided an up-to-date rs ID synonym file for the most recent dbSNP release
  - included filtered versions of this synonym file with the 1,000 Genomes reference data files (will be auto-detected)
- added additional options for dealing with duplicate SNP IDs (including synonyms) in --pval input file
  - changed default behaviour from terminating with an error to removing duplicate SNPs
- added option to condition gene analysis on specified SNPs in the data
- added --big-data option, for processing very large data sets much more quickly and using much less memory
  - added auto-detection, turning on --big-data automatically for sample sizes greater than 25,000
- modified chromosome batch mode to significantly reduce memory usage per batch even when reading from a single genotype data file
- changed NPARAM for the SNP-wise (mean) model to align more closely with the NPARAM for the PC regression model
- fixed minor bug: in v1.05, the column names P_MAIN and P_INTERACT for the interaction model were switched (fixed in v1.05b)

v1.05
- added optimized top1-SNP p-value gene analysis model
- added multi-models: running multiple gene analysis models (PC regression, SNP-wise (mean), SNP-wise (top 1) for each gene and combining into a single p-value
- changed PC regression gene analysis model from PCA on SNP covariance matrix to PCA on SNP correlation matrix (more balanced in presence of rare variants)
- deprecated --snp-wise flag, replaced by 'snp-wise' modifier of --gene-model flag
- added gene by covariate interaction model

- added multi-burden score mechanism, setting a (modifiable) maximum on the number of variants aggregated into a single burden score (creating multiple burden scores if exceeded)
- added automatic burden score mechanism for very low MAC SNPs (SNPs with both MAC <= 25 *and* MAF <= 0.01)
  - turned on by default, can be turned off if desired

- added inverse mean MAC per gene (and log value) as covariate in gene-level analysis, to correct for low power in genes containing mostly very low MAC SNPs
  - added option to selectively include automatic covariates
- changed default imputation of missing gene covariate values from mean imputation to median imputation
- added option to perform two-sided test for competitive gene-set analysis

- added a batch mode for distributed/parallel analysis of larger data sets
  - added chromosome batch mode and support for input data split by chromosome

- performed major revision and expansion of the user manual

v1.04
- removed single imputation for missing genotypes, replaced with analysis with rescaled sufficient statistics computed from observed data
- added permutation-based empirical multiple testing correction for gene-set/gene property analysis
- renamed --rare to --burden
- replaced --burden SNP filter-file options with general SNP filter-file options under --gene-settings
- added R-squared and adjusted R-squared values to gene analysis for PC regression output (default model when performing raw data analysis)
- added regression coefficients (raw and semi-standardized) and standard error in output for competitive gene-set analysis and for gene property analysis
- turned off self-contained gene-set analysis by default (can be turned on with 'self-contained' modifier for --set-annot flag)
- added option to filter genes used in gene-set / gene property analysis based on filter file

v1.03
- added permutation based multiple-testing corrected p-values for gene-level analyses
- added option to override sample size to meta-analysis
- added SNP differential missingness filter
- expanded functionality for gene property analysis (handling of missing values, conditional analysis)
- expanded functionality for conditional analysis (increased maximum conditioned on sets/covariates, added gene set specification through file)

v1.02
- additional rare variant analysis options
- added meta-analysis options

v1.01
- added support for nonhuman genomes
- stricter default pruning of genotype data (raw data PC regression model) to improve power
- minor adjustments to computation of gene correlations to improve type 1 error rate control in gene-set analysis