*** v. 1.8-0
(2013.12.06)
Fixed memory leak in a number of functions and deleted unused variables.
Moved data objects ge03d2.clean, ge03d2c, ge03d2ex.clean, ge03d2ex, ge03d2 and srdta to a specially created new R package called GenABEL.data.
*** v. 1.7-7
(2013.11.04)
Fixed bug #5040: Spelling of the name of G. Svischeva incorrect.
(2013.08.27)
Removed CRAN version checks following the request from B. Ripley,
fixing other problems identified by --as-cran checks ("iterator"
package changed from "DatABEL" to "GenABEL"; renaming CHANGES.LOG to
ChangeLog), releasing this version on CRAN
(2013.07.19)
Added warning that *2databel's may not produce output in desired data type
and fall onto float
(2013.06.29)
Changed the 'polygenic' nullification threshold to 1e-8
Added references for Grammar+ transformation
Allowed for user-specified output data type in '*2databel' functions
Changed version number to 1.7-7
(2013.06.06)
Two new functions added: LiLog() and palinear2LiLog() for the
estimation of logistic beta's from a linear regression. Thanks to
Nicola Pirastu for this contribution. The paper that describes these
methods has been submitted (see the help of both functions for the
citation).
(2013.06.03)
Fixed bug #2664 in export.merlin() (exported files differ when using
different settings for the stepids argument).
(2013.05.30)
Fixed memory leak in export.plink() when exporting to .ped format (Bug #2055).
*** v. 1.7-6
(2013.05.16)
Negative eigenvalues set to 'very small' (1e-16) instead using abs
in 'polygenic': this is more justified
Fixed too long lines in Rd files.
(2013.05.15)
Updated version number to 1.7-6 (release), checked with latest
dev-version of R available. Added 'bigRR' as suggested package.
*** v. 1.7-5
(2013.04.09)
Fixing bug [#2672] (GenABEL::cocohet default graph output). Now by
default the function does not produce graphics (makePlot = FALSE); if
graphics is requested (makePlot = TRUE) it is produced using a default
plot device (so a user can redirect the output to a file if (s)he
desires and in the way (s)he desires).
(2013.04.04)
Fixing bug [#2525] (http://r-forge.r-project.org/tracker/index.php?func=detail&aid=2525&group_id=505&atid=2058)
Thanks to Vladimir Naumov for submitting the patch!
Updated documentation files
(2013.04.03)
Added patch to bug [#1287] - example(ibs) producing strange clusters.
Thanks to Vladimir Naumov for submitting the patch!
(2013.04.03)
Upgraded the version number to 1.7-5
(2013.04.02)
Changed default behaviour of the export.plink() version. Export to .tped is now
the default (because of the bugginess of the .ped export, see bug #2055.
*** v. 1.7-4 (2013.02.22)
(2013.02.22)
replacing SNP with get("SNP") in impute2databel with option 'old' so to avoid
CRAN checks complains
(2013.02.03)
Fixing bug in 'grammar': when using 'gamma', effects and s.e.s are not re-computed
(bug reported by Xia Shen)
(2013.01.18)
Fixing bug in ztransform (if NA is present and formula used, NA's are kicked out)
Changed the version number
*** v. 1.7-3 (2013.01.09)
(2013.01.09)
Commented the parts related to non-additive GC in qtscore
Removed calls to 'attach' from multiple procedures
Decrease of running time for long-running examples (GC_ovdom,GC,check.marker,Xfix,srdta)
(2013.01.07)
Fixing the problem which prevents the package from loading while checking the version on CRAN
(2012.12.03, YA)
- Changes in DESCRIPTION and version to denote new working
version
- added argument 'eigenOfRel' to 'polygenic' allowing passing
the results of 'eigen' of the relationship matrix instead of the matrix
itself (makes much faster evaluation). See unit test test.polygenic.eigenOfRel
for use example
- found out that impute2databel is broken, see bug report [#2418]
*** v. 1.7-2 (2012.10.14)
(2012.09.14, YA)
Changes in DESCRIPTION and version
(2012.08.29, YA)
Small changes related to CRAN release of 1.7-1; this release is to be
updated to 1.7-2 as soon as the Grammar-Gamma paper is published and
some feedback is collected for the 1.7-1
(2012.08.15, YA)
Small updates to documentation, verifying that checks are passed.
Many NOTEs about use of partial arguments - something to fix
eventually.
(2012.08.09, YT)
Add new function GC (Genomic control for non-additive models) and GC_ovdom fot overdominant model.
Update some functions.
(2012.07.06, YT)
Removed documentation file grammar.old.Rd as obsolete
Add new function PGC (Polynomial genomic control for non-additive models)
*** v. 1.7-1 (2012.04.27)
Replaced 'grammar' function with new one, allowing 'raw', 'gc'
and 'gamma' varieties of the method.
Added option weight = "eVar" to 'ibs'. This uses empirical variance of
genotypes when estimating kinship matrix. This is useful when working with
such data as Arabidopsis.
*** v. 1.7-0 (2011.12.20)
Addressed bug [#1383] (message about upgrade on loading the library
not entirely correct) by adding new function checkPackageVersionOnCRAN()
and re-writing zzz.R (.onLoad) using it.
Addressed bug [#1673] (bug in load.gwaa.data and export.merlin when
sorting is enabled, filed in by Daniel Taliun), added unit test
runit.sortmap.internal.R/test.sortmap.internal.bug1673
Addressed bug [#1676] (regression in merge of gwaa.data, phenotypes
don't get merged correctly; filed in by Lennart Karssen), added unit
test runit.merge.R/test.merge.bug1676
Added 'KS' method to estimate lambda in 'estlambda'. Generates rather
good results (under null) and may be considered as default option in the future,
after testing under the alternative.
Account for situation when HGLM fails to converge (s.e.'s set to NA in
h2$h2an$se)
added option 'transposed' to 'export.plink' to export TPED files
speeding-up 'mmscore' by x1.8 through more efficient vector-matrix
product computations
export.merlin, export.plink now runs much faster (exports on C++ code)
Bug [#1641] (regression bug with merge.snp.data in version 1.6.9;
filed in by Karl Froner) fixed. Now GenABEL deals with exceptional
situation in merge.snp.data / monomorphic part; added case of no overlap
in SNPs (skip monomorphic stuff then). RUnit regression test
runit.merge.R/test.merge.bug1641 added.
Modifications in 'estlambda': plot=FALSE by default, added option
to estimate Lambda with median method; added option 'filter' for
filtering SNPs with 0-statistics
*** v. 1.6-9 (2011.08.30)
Change the name of 'maintainer' to genabel.project@gmail.com to avoid spamming
CRAN people
*** v. 1.6-8 (2011.08.25)
added 'strand<-' method for snp.data and gwaa.daat-class
added 'id' argument to 'load.gwaa.daat'. This may be handy when eg
importing PLINK phenotypic data id="IID"
Update of 'polygenic_hglm' by Xia Shen. Quote:
We've updated the hglm package to version 1.2-2 recently on CRAN, where
the major updates include:
- Sparse matrix implementation
- Multiple random effects
- Function hglm2() which accepts lme4-style formula input
In this update of polygenic_hglm() the speed could be slightly faster than
before since the hglm() now uses sparse matrix technique (however I don't
expect it to be much faster since the kinship matrix is always dense). Other
modifications are:
- The current update of hglm package does not spit out profile log-likelihood
since we haven't perfectly solved the likelihood computation for multiple random
effects yet. But since polygenic_hglm() only needs one random effect (polygenic
effect), I've re-implemented the likelihood computation directly in
polygenic_hglm(), and no need to specify 'method = "REML"' anymore.
- In order to solve the problem from the forum about inputting formula as
'y' or 'y ~ 1', I've added a check before creating the model.frame, so now
both types of inputs should work.
Added 'recodeChromosome' function, which must be handy when e.g. importing data
from other software which uses integer for sex chromosomes and other non-autosomes
(X, Y, mt).
+> tutorial genetic data QC - list all options, do not use default?
'check.marker' added "none" (equal to ibs.mrk = -1); set "both" to default
Updated documentation for 'scan.glm.2D'.
Added GRAMMAR+ transformation and computation of GRAMMAR+
correction factors to 'polygenic'.
The default value for the stand option to convert.snp.tped() is now "u",
instead of "+". Now convert.snp.tped() and convert.snp.ped() have the
same defaults.
Added patch of bug [#1322] + regression test (contributed
by Nicola Pirastu, see
https://lists.r-forge.r-project.org/pipermail/genabel-devel/2011-May/000276.html).
This is still suboptimal treatment -- the covariate is just dropped;
in say 'lm' it is 'kept' with NA as estimate, and IDs with missing
data dropped, which is not the case with 'polygenic'.
Speeding up 'polygenic' by avoiding multiple inverses of
the matrix.
Fixed bug in 'polygenic' causing the offset in
names(...$residualY) and names(...$pgresidualY) in
case of missing observations.
Updated the merge.snp.data procedure to allow meaningful
merging of mono- and poly- codings. Take care -- this is
new and not tested much yet! You best check few results
manually after the merge!
Added coding<- method to gwaa.data and snp.data classes
Added I/D encoding and monomorphic classes (AA, TT, GG, CC,
--, II, DD, ...) to 'snp.data' and convert.snp.*
Updated 'polygenic' with use of 'polylik_eigen' developed by
Gulnara Svischeva. Now 'polygenic' works MUCH faster. The
main advantage of Gulnara's method is that time to compute
the likelihood function is approximately linear with number
of subjects. In that, relative speed-up grows with sample
size growth, e.g. for 100 IDs, it is x15, x40 for 200,
x60 for 400, and xX for 800 individuals (using two fixed effect
covariates, see runit.polylik.R). Also added RUnit test to check
consistency of results based on old 'polylik' and new
'polylik_eigen'.
Upgrade version number
*** v. 1.6-7 (2011.05.17)
Submitted 1.6-7, based on r727, to CRAN
Deleting some pdf from 'doc' and compressing 'data'
Fixed 'technical' bug [1398] (related to changes in R 2.14):
https://r-forge.r-project.org/tracker/index.php?func=detail&aid=1398&group_id=505&atid=2058
Set 'GenABEL developers' as a maintainer of the package;
set the genabel-devel list as 'maintainer' e-mail
Fixing bug [1388], added regression test
https://r-forge.r-project.org/tracker/index.php?func=detail&aid=1388&group_id=505&atid=2058
upgrade version number
fix in example of polygenic_hglm
*** v. 1.6-6 (2011.04.29)
Added Xia Shen's procedure 'polygenic_hglm'. Features: quick
convergence, standard errors for fixed effects.
Updated check.marker, added warning if no. Y-markers is < 10
Applied the patch of Nicola Pirastu
http://lists.r-forge.r-project.org/pipermail/genabel-devel/2011-March/000182.html
to descriptives.trait. Added RUnit regression tests, updated
documentation. Bugs fixed: [#1184], [#1185], [#1259]
Added 'reconstructNPs' function; added extra options
to 'findRelatives'.
Added a number of functions facilitating relationship checks.
The core function is 'findRelatives'. Compared
to guessing relations from genomic kinship matrix, this
procedure offers several enhancements: (1) by use of IBD/IBS
3-state space, it allows to distinguish between some pairs,
which have the same kinship (e.g. parent-offspring from
brother-sister; uncle-nephew from grandparent-grandchild, etc.)
(2) it reports likelihood, allowing for more rigorous inferences
Changes in convert.snp.mach documentation to reflect the fact that
the map-file should have header; changes to convert.snp.illumina
documentation to reflect the nature of the data better; added
option 'mapHasHeaderLine' to convert.snp.ped and updated documentation
(resolving feature request #1317).
***** v. 1.6-5 (2011.02.07)
Added '#include <cstdarg>' to iterator.cpp to solve
'[#1273] GenABEL 1.6-5 does not compile'
commented S4methods... in NAMESPACE to meet the
request from the R-team; patched documentation for
summary.scan.gwaa and summary.gwaa.data
in order to avoid NOTE about S3methods "usage"
section
added extra checks for impute2databel to ensure that
sample file header is not used
change in merge.snp.data by Maksim Struchalin, adding option
"intersected_snps_only" (if TRUE, then only interesected SNPs
will be in output)
Disabled check on intermediateXF in check.marker by setting intermediateXF=c(.5,.5)
by default. This will be dealt with further (see tracker '[#1210] intermediate
inbreeding checks with check.marker' )
patched Roxygen documentation for qtscore
use of 'setTxtProgressBar' in qtscore with times > 1 to indicate progress
polygenic() added default option patchBasedOnFGLS, if no convergence based
on difFGLS criterion, fixed effect betas are patched to FGLS betas before
starting re-iteration
adding a wrapper chi2_CG -> cocohet
fixed 'demo(ge03d2)'
'stepids' argument added to export.merlin
Fan/Maksim's procedure for cocohet (CCH) procedure added, renamed by Yurii
Bug fix in export.plink() thanks to Nadya Belonogova
Made sure that SE reported does not carry the sign of beta (always 'abs')
added mechanism to check for updates at load time
moved summary.snp.data to 'iterator' framework, changes in gwaa.data@gtdata@gtps
definition, now 'ANY' allowing for use of different types of data with
interator. Interestingly, for 'old' data type computations became faster!
***** v. 1.6-4 (2010.08.30)
new official release tested by ESP29-2010 participants
many thanks to everybody who provided feedback!
***** v. 1.6-3 (2010.08.14)
changes in the code and tutorial due to preparing
for ESP29 course
added function 'export.plink'
added multiple methods to work with 'gwaa.data', 'snp.data' and
'scan.gwaa classes, such as 'annotation', 'coding', 'refallele',
'effallele', ...
more annotations are delivered for summary objects
***** v. 1.6-2 (2010.08.07)
changes due to DatABEL functions names change
***** v. 1.6-0 (2010.06.21)
changes in DatABEL-embed functions used by impute2databel (treating of NAs);
added tests for that situation
***** v. 1.5-9 (2010.06.18)
change in DESCRIPTION anyDuplicated() is in R 2.10.0, but not in 2.4.0
small changes in polygenic -- convergence checks with intercept-only model; treatment
for h2-boundary-stick situations (at small N)
***** v. 1.5-8 (2010.06.10)
multiple changes in polygenic default parameters
and extra criteria now allowing better
convergence. Many thanks to Nicola Pirastu for spotting
poor convergence when multiple parameters were estimated!
added methods "dim" and "dimnames" to snp.data-class
***** v. 1.5-6 (2010.05.28)
fixes in filevector
Many thanks to Pau Navarro for help in identification of the bug!
***** v. 1.5-4 (2010.05.09)
changes in filevector related to DatABEL not passing checks on Windows
(disconnect used FiltredMatrix destructor, which does not call AbstractMatrix destructor)
***** v. 1.5-2 (2010.05.04)
switching to new DatABEL with single class
***** v. 1.5-1 (2010.03.17)
further improvement and bug fixes in databel-facilitated functions
***** v. 1.5-0 (2010.03.17)
update of ABEL-tutorial (also few details on NCBI2R added)
general clean-up, submission to CRAN
extended add.phdata()
added del.phdata(), shortcut to @male slot (function male())
added checks for F in sex checks, ...$otherSexErr lists individuals with 0.2 > F > 0.8
multiple changes to xxx2xxx (mach, impute, databel)
***** v. 1.4-9 (2010.02.06)
added arrange_probabel_phe(), also shortcuts to
slots of gwaa.data-class: chromosome, coding,
gtdata, idnames, map, nids, nsnps, phdata, snpnames, strand
***** v. 1.4-8 (2010.01.24)
added export.impute(), patch_strand(), mach2databel(), impute2databel(),
impute2mach, extract.annotation.impute(), extract.annotation.mach()
functions
***** v. 1.4-7 (2009.12.03)
cleaned help for polygenic()
Diagonal elements returned by ibs(...,w="freq") changed from (0.5+F) to
0.5*(1+F) (thanks to Chris Haley for pointing this problem out!)
***** v. 1.4-6 (2009.11.06)
Default formetascore's argument transform set to "no"
***** v. 1.4-5 (2009.09.09)
New function "var.meta" added.
Function is intended for pooling (meta-analysing) trait means for
different genotypes separately.Input files have to be like in example
below (plink generate this kind of files). At the end you have one file
in same format but with pooled MEANs and SDs
***** v. 1.4-4 (2009.09.02)
bug fix in hom() and perid.summary() -- only polymorphic markers were used to
report call rate etc. Many thanks to Patrice Godard, who has reported the bug!
The bug only appeared in version 1.4-3
project submitted to r-forge
***** v. 1.4-3 (2009.07.21)
qvalue moved to "suggests"; no strict dependency on availability of that anymore
(thanks to Beate Glaser for noting that tcltk is not so obvious component
of all systems)
changes to C++ LM code to improve portability (Solaris 10)
hom() function updated, now producing variances (to be used as diagonal
elements of the genomic kinshp matrix when doing EIGENSTRAT analysis).
Many thanks to John Barnard for contributing the code!
all class checks changed to ensure compatibility with S4
(thanks to Maxime Rotival)
***** v. 1.4-2b (2009.02.20)
small fix in merge.gwaa.data (by="id")
***** v. 1.4-2 (2009.01.28)
Changes in polygenic()/mmscore() to allow for sub-setting on
IDs
Bug fix in mmscore(). The bug concernes estimates of beta, the
P-values reported by mmscore() were not affected.
***** v. 1.4-1 (2008.11.26)
merge.gwaa.data, merge.snp.data well tested and
recommended for wider use (special thanks to Arne
Schillert and Michael Preuss)
Modifications of some functions for better parallelizability.
[user-invisible] changes to: ibs, hom, r2fast, perid.summary
(thanks to Surakameth Mahasirimongkol and Unitsa Sangket)
Resolved R 2.8.0 issues with load.gwaa.data.
(thanks to many people reporting the problem)
Partly resolving incompatibility between the way GenABEL
computes the diagonal elemnts of the "covariance matrix"
for use in egscore() and the original EIGENSTRAT matrix
(many thanks to Guiseppe Palermo)
***** v 1.4-0e (2008.09.18)
Small changes in mlreg for consistency to formetascore required input
Bug fix in mlreg (would produce meaningful results when no. snps is
dividable by 4)
***** v 1.4-0d (2008.08.21)
bug fix in merge.snp.data() -- many thanks to Arne Schillert and Michael Preuss
***** v 1.4-0c (2008.07.25)
bug fix in estlambda() -- many thanks to Franz Quehenberger
***** v 1.4-0b (2008.07.21)
Multiple fixes in merge.snp.data
Fix in check.marker (for Y-chromosome, the call rate
was estimated using all people)
***** v 1.3-9 (2008.04.23)
Function mlreg() (aka reg.gwaa()) added. Implements maximum likelihood
estimation and Wald test for linear and logistic regression and Cox
proportional hazards models.
Function convert.snp.affymetrix() added
***** v 1.3-7 (2008.04.16)
Function merge.gwaa.data() added
Modifications to formetascore() (now can run with mmscore())
Few cleanups
***** v 1.3-6 (2008.03.20)
Function merge.snp.data() added
npsubtreated() updated with new functionality
"quality" measure used in convert.snp.mach changed to 7th column (r2)
plot.scan.gwaa(): option df="all" removed, options col, ystart added
add.plot(): option colors added
Improvement of formetascore function
***** v 1.3-5c (2008.02.27)
Bug fix in check.maker.internal (with "lower" option, would not exclude
on IBS).
Small bug fix estlambda (would not run if supplied with P-values, and
some P-values are < 1.e-16). This bug did not affect any other function.
***** v 1.3-5b (2008.02.21)
Bug fix in ztransform(): would not work without data argument,
if formula supplied will not use it. This bug also affected rntransform().
***** v 1.3-5 (2008.02.15)
qtscore "binomial" workout
minor bug fix in check.marker() -- the procedure would fail to run
in presence of more than couples of "twin" (e.g. multiple plant clones).
bug fix in formetascore() function: standard errors were reported with
sign of beta
***** v 1.3-3d (2008.02.11)
A fix in npsubtreated() function (in some situations all medicated
people's values were set to NA)
***** v 1.3-3c (2008.02.03)
npsubtreated() function, to be used in analysis of
traits affetcted by treatment (medication) added.
formetascore() procedure extended with parameter "verbosity".
***** v 1.3-3 (2008.01.12)
autosomal() function added
Y and mtDNA markers-specific analyses added in QC function check.marker()
This also involves checks for people with XXY genotypes.
Robust Genomic Control added to correct 2df test in qtscore
function -- yet experimental.
formetascore() function introduced -- analysis oriented to
future meta-analysis.
ztransform function created, rntransform function modified to take
similar syntax as ztransform.
Bug in convert.snp.merlin (similar to the one in convert.snp.illumina)
fixed.
***** v 1.3-2 (2007.12.18)
Extensive cleanup, including fixes to mmscore and grammar;
examples introduced for polygenic, mmscore, grammar, and others.
Bug in convert.snp.tped (similar to the one in convert.snp.illumina)
fixed.
The number of people/genotypes (N) used in analysis of a particular SNP
added to the output of qtscore, mmscore, ccfast, egscore, and grammar.
plot.scan.gwaa changed -- now it is possible to use ylim.
Bug fix in qtscore with binary traits without covariates -- there was
no test whether coding is 0/1 (if not, test would be wrong).
Changes in C code of concerns functions convert.snp.mach and
convert.snp.ped. Speed of conversion from MACH format greatly
improved.
***** v 1.3-1 (2007.11.05)
The strand information as based on build 35, is available from Yurii
for Illumina 318K.
Bug fix in convert.snp.illumina (alleles would change to the other one
in large proportion of SNPs, e.g. A/G polymorphism, AA -> GG, GG -> AA).
Though this does not affect results directly (e.g. P-values and the absolute
effect values are OK), this may couse confusion when you compare your
results to other people's results (the sign of the effect may be swapped),
also imputations would have gone wrong.
Small bug fixes with strand operations (SNP names were lost in sub-setting).
New procedure rntransform(). This procedure performs
rank transformation to normality. The transformed variable
is perfectly normal, unless there are ties in the original
distribution.
***** v 1.2-9 (2007.10.22)
A bug fix in check.markers(ibs.exclude="lower") -- before the fix, the
samples with HIGHER call were excluded. Thanks to Tuuli Lappalainen!
Bug fix in refresh.gwaa.data
option "details" added to qtscore().
When details=FALSE, SNP and ID names
are not reported, which saves some memory
***** v 1.2-8 (2007.08.27)
GenABEL tutorial updated and, internals joined with
tutorial, new introductory sections added.
New association analysis procedure introduced.
egscore() computes association statistics adjusted
for eigenvector(s), following Price et al, Nat. Genet.,
2006.
Improved and new functionality of check.marker():
much faster X-errors check, identification of female
which are likely to be male. When "twin" DNAs are
identified, the one with lower call may be dropped,
or both (previously random).
***** v 1.2-7 (2007.08.13)
New list names added to output produced by qtscore(),
grammar(), ccfast() and mmscore(): chi2.1df and
chi2.2df. These correspond to Chi-square values produced.
The minimal P-values, as before, are resti=ricted to 1e-16
Improvements in plot.scan.gwaa() and add.plot() --
now 0-probabilities converted to 1.e-16 (previously
a warning message would be produced and the point would be
dropped); also option df="Pc1df" introduced.
***** v 1.2-6 (2007.08.08)
A major improvement in GenABEL ability to import and export data
compatible in format to other packages and databases.
NOTE: to convert your R data from old to new format within R, use
refresh.gwaa.data() function. Old phenotype- and genotype- files will load
with load.gwaa.data with no problem.
One of the main goals behind this effort was to make GW imputations
with MACH software (G. Abecasis) easy.
(a) New format of GenABEL data (0.1) developed. Now the
features stored by GenABEL include SNP coding (actual nucleotides)
and strand.
(b) New and updated convert.snp.* functions, which convert genotypic
data from external to GenABEL format
-- convert.snp.ped: large improvement in speed (now in C++),
flexible control. Pre-makeped, MERLIN
and MACH formats supported
-- convert.snp.tped: some bug fixes, switch to 0.1 data format
-- convert.snp.mach (NEW!): converts MACH files to GenABEL
format, allowing for filtering on
quality
-- convert.snp.illumina (NEW!): converter for Affymetrix / Illumina
style of files
(c) export.merlin() function to export data in MERLIN format. MACH
can be directly applied on these files. Coupled with
convert.snp.ped(format="mach") this allows easy data export,
imputations, and re-import.
***** v 1.2-5 (2007.07.11)
Some improvement of convert.snp.ped -- now is able to read
alleles coded as characters
***** v 1.2-4 (2007.07.10)
Few small fixes done for compatibility with R 2.6.0-devel
convert.snp.tped() procedure by Toby Johnson added.
This procedure facilitates fast and effective conversion
from PLINK TPED ("transposed pedigree") format to the
GenABEL format. See help(convert.snp.tped).
Few other small improvements. This is beta release version
for procedures facilitating analyses using genomic
kinship (IBD) matrix, such as polygenic(), mmscore() and
grammar(). We tested these procedures with quantitative traits
and they worked fine; for binary traits there may be few
bugs yet.
***** v 1.2-2 (2007.06.07)
Improvement in polygenic() to ensure convergence, two
optimisation methods are now allowed.
"Formula" part in qtscore changed to more standard format
(now one can use any terms allowed in glm, no CRSNP term
needed).
Bugs in grammar and mmscore fixed. Previous version tended
to produse some very wrong results.
Polygenic silightly modified (new output objects added).
A fix in polygenic with covariates.
Other bug fix in stratified analysis with X-chromosome.
***** v 1.2-1 (2007.06.04)
A set of new procedures to do analysis in samples of related
individuals. They also can be used for analysis of samples
with population staratification and cryptic relations.
polygenic() provides means to estimate "heritability" and effects
of covariates, in data having some correlation, as described by
(genomic) kinship matrix.
Mixed-model score test mmscore() added. This implements the test
for association between the vector of phenotypes y and vector of genotypes
g by computing
((y-E[y]) Inverse[V] (g-E[g])^2/((g-E[g]) Inverse[V] (g-E[g])
where Inverse[V] is the inverse of variance-covariance matrix
and E[y] is expectation of the trait, as provided by polygenic().
grammar() procedure implements fast variant of the above score test
((y-E[y]) Inverse[V] (g-E[g])^2/((g-E[g]) (g-E[g])
This statistics does not have chi-squared distribution, but becomes
distributed so after GC (thus Pc1df are fine P-values to use)
NOTE: grammar is still experimental.
An error in qtscore, which led to using always only 2 strata
(male/female) in analysis of X-chromosome fixed.
***** v 1.2-0 (2007.04.17)
Large improvement in the speed of the qvaluebh95() procedure.
The time used to grow quadratic with the number of P-values to
evaluate, now reduced to linear.
ibs() procedure "freq" option is updated to provide weighted
measure of genomic identity.
dprfast() facilitating fast D' computations added. The results
produced are not exacly the same with D's from LD/genetics.
***** v 1.1-9 (2007.04.04)
A procedure for fast r2 computations (r2fast()) added.
It seems that this procedure results in r2s which are biased upward
at small values, compared to r2s from LD/genetics.
A procedure to estimate Morton-Collins rho added (rhofast())
***** v 1.1-8 (2007.03.12)
empirical derivation of GW significance is now embedded into
qtscore/ccfast, emp.* are now serving as shortcuts to
call these functions with argument times=100
qtscore: changes to treat binomial variables (residuals from glm)
Added new data sets: ge03d2, ge03d2c
Added demo ge03d2; this ones shows extensive QC + how to deal
with genetic stratification.
***** v 1.1-7 (2007.03.05)
In scan.haplo.2D procedure with simulate = TRUE change 0s -> 1/(N+1)
add.plot.scan.gwaa() function changed to add.plot(), now delas with
"scan.gwaa" and "scan.gwaa.2D" objects
as.data.frame.gwaa.data which takes out @phdata part is arranged
Fixed bug which made qtscore, ccfast and emp.* use phenotypes from
the attached data frame, even when the data argument was supplied.
For qtscore and emp.qtscore Formula now can be formulated as character
and a formula.
hom() function developed (previously heteroz; now is implemented in C)
perid.summary() designed to produce average heterozygosity and call rate
per person; this implied changes in check.marker() and summary.check.marker()
ibs() function developed, this one extends to check.marker()
A family of descriptives.* functions added: .marker, .trait, .scan
Much improved genetic data QC (check.marker); new function Xfix()
***** v 1.1-6 (2007.02.12)
Improved memory managmenet. In particular, @chromosome slot is now
stored as a factor.
***** v 1.1-4 (2006.12.21)
Due to accumulated changes, version number updated.
Multiple improvements and fixes + catable() function
***** v 1.1.3 (2006.11.12)
scan.glm.2D and scan.haplo.2D functions added
run.haplo.score.slide renamed to scan.haplo for consistency
heteroz() added
scan.gwaa-class is modified: effects are also returned.
This led to modification of C code for fastcc, qtscore, R
code for qtscore(), ccfast(), fcc(), emp.qtscore(), emp.ccfast(),
scan.glm()
improved memory management in functions ccfats(), emp.ccfast(),
qtscore(), emp.qtscore(), summary()
show.ncbi() added
***** v 1.1.2
scan.glm patched (it did not work with df=2 when a SNP had only 2 genotypes)
minor fix in sub-setting of snp.data class
ccfast, qtscore, scan.glm, emp.ccfast, emp.qtscore and plot.scan.gwaa are modified.
Major change concerns ccfast and qtscore: now, 1-df and 2-df tests are
computed simultaneously within the C functions.
The class "scan.gwaa" now contains both 1- and 2-df P-values.
This also addresses the point on TODO:
2006.08.17 implement qtscore function with 2df
Small fix in qtscore and ccfast (to act on 1 d.f. when # genotypic classes is 2)
The function scan.gwaa removed as redundant
qtscore now allows for analysis using strata (Cochran-Mantel-Haenszel)
Added average/median chi2 to the "scan.gwaa" class (to be used for genomic
control)
TODO fixed:
2006.10.11 q-value returned by qvaluebh95() is wrong
emp.qtscore and emp.qtfast optimised (runs 3-10 times faster) by
removing unnecessary checks and calling C function directly
(instead of calling ccfast/qtscore)
NAMESPACE introduced
Update of documentation: now R CMD check passes all checks
Formal class introduced for scan.gwaa-class
demo(srdta): demo(srdtawin) is made by removing load-data from demo(srdta)
because this may not work when there are no administrator's priviliges
last modified 08.11.2006
***** v 1.0.2
Function scan.glm is introduced; this allows to scan for interactions in
an easy way. This is essentially a modified scan.gwaa.
New function to convert ped-files type of data to internal format,
convert.snp.ped, was arranged
Function add.plot.scan.gwaa was added
Library "methods" loaded on start-up
last modified 01.11.2006
***** v. 1.0.1
is resolving the issue of
2006.07.20 need fast exact test for HWE, along the lines of fastcc,
as HWE.exact is too slow (Yurii)
and
2006.10.11 gwaa.data should be easily sub-settable -- now one
needs to extract @gtdata and subset and @phdata and
subset and then merge (also using base R things) (Yurii)
Generated by dwww version 1.15 on Sat May 18 09:42:47 CEST 2024.