The hardware and bandwidth for this mirror is donated by METANET, the Webhosting and Full Service-Cloud Provider.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]metanet.ch.
m_causal=1 bug (used to die, now works
correctly).sigmaSq to sigma_sq
(parameter of sim_trait and cov_trait).sim_trait return list has more descriptive names:
trait (old y)causal_indexes (old i)causal_coeffs (old beta)popkin 1.2.0.9000
changes.README changes in parallel with
bnpsd 1.1.0.9000 changes.allele_freqs, which handles
BEDMatrix objects correctly. Rest of sim_trait
doesn’t handle BEDMatrix objects yet.sim_trait, and consequently the whole package,
supports BEDMatrix objects. This is ideal for simulating
traits from real and potentially very large genotype data.allele_freqs to load genotype chunks from a
BEDMatrix object more efficiently, using as much available
memory as possible. Requires an updated popkin package
(>= 1.2.6.9000).class usage now that matrices
return a two-element array in R-develman/figures/sim_trait changed default maf_cut
from 0.05 to NA, and in that case code no longer computes
marginal allele frequencies unless they are needed because
p_anc is missing (and kinship is provided).
This is much faster and memory efficient in extremely large simulations,
and just makes more sense to me generally (the MAF threshold was a
quirky feature, still available but non-default now).sim_trait now delays calculating allele
frequencies if maf_cut = NA and p_anc is
missing (and kinship is provided), so that these
frequencies are only calculated on the small subset of loci selected to
be causal (rather than the whole genome, which was the original
behavior). This is expected to speed-up trait simulations from real
genotypes.sim_trait_mvn, which draws traits from
the multivariate normal distribution with covariance structure matching
that of genetic traits (also called “infinitesimal” model). These traits
are useful for testing heritability estimation. A visual validation of
these simulated traits is available in the vignette.alelle_freq was modified to reduce memory
usage for the BEDMatrix case.
sim_trait, the
BEDMatrix-specific options mem_factor and
mem_lim were replaced by the option
m_chunk_max.popkin is no longer a
dependency of this simtrait package. However,
popkin is still recommended for estimating the kinship
matrices required by some of the functions of this simtrait
package.simtrait:
rmsd: General root-mean-square deviation (RMSD)
functionpval_srmsd: Signed RMSD between observed and expected
(uniform) p-valuespval_aucpr: Area under the curve (AUC) of
precision-recall (PR) curve
PRROC package dependencypval_infl (classic inflation factor, but
from p-values) to complement yesterday’s new functionsallele_freqs added option fold,
which if TRUE returns folded i.e. minor
allele frequencies. Default is FALSE, to return allele
frequencies for the alternative allele (the allele counted as it given
in the genotype matrix, whether it is the minor or major allele).sim_trait added option
const_herit_loci, which when TRUE constructs
causal coefficients as inversely proportional to the square root of
p*(1-p), where p is the ancestral allele
frequency. This ensures equal per-causal-locus contribution to trait
variance. Default draws causal coefficients randomly from a standard
normal distribution, rescaled to result in the desired heritability, for
unequal per-locus contribution to trait variance.herit_loci, which calculates per-locus
heritabilities based on variance formula (in terms of allele
frequencies, coefficients, and overall trait variance factor).
sim_trait in unit tests.sim_trait with option
const_herit_loci = TRUE now adds random signs (+/-) to the
causal coefficients.
sim_trait now requires that p_anc
have the same length as the number of loci in X (stops with
an error otherwise). Previously this was not checked and could return
traits that were NA for all individuals without clear
indications that anything was wrong.pval_srmsd now accepts
causal_indexes = NULL to handle cases where all p-values
are null (before a NULL input would cause an error).sim_trait renamed option
const_herit_loci to fes (fixed effect sizes),
the language used in the paper.NEWS.md slightly to improve its
automatic parsing.LazyData: true (to avoid a new “NOTE” on
CRAN).URL and
BugReports)par() in vignette examples.sim_trait fixed an important bug resulting in
misspecified heritability!
rnorm the desired variance
(1 - herit) * sigma_sq where the standard deviation (its
square root) was required!herit value by
herit / ( herit + (1-herit)^2 * sigma_sq ). If
sigma_sq = 1 (default), the effective heritability was
always larger than desired!sim_trait bug and its fix.inbr_diag in plot_popkin calls,
which in this case made diagonal values larger (as they were larger than
1), among other minor adjustments.sim_trait, sim_trait_mvn, and
cov_trait: added parameters labs and
labs_sigma_sq to simulate/model non-genetic group effects.
sim_trait also now returns
group_effects as one of the named elements of the return
list.sim_trait parameters p_anc and
kinship now have default NULL values (used to
not have default values), to facilitate scripting in cases that can be
either simulated or real genotypes.pval_type_1_err and
pval_power_calib for calculating type I error rates and
calibrated power, respectively.pval_type_1_err and
pval_power_calib:
alpha can be a vector, and return value is a
vector of estimates, one for each alpha.1/alpha.pval_type_1_err and
pval_power_calib:
pvals is NULL and
alpha had more than one value, these functions incorrectly
returned a scalar NA, now they return a vector with
NAs the length of alpha.pval_infl if input pvals = NULL
now returns NA (used to return
numeric(0)).sim_trait, sim_trait_mvn,
cov_trait:
check_labs renamed to
check_herit_labs and:
herit is now firstherit checks that used to be in upstream
functions into hereThese binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.