variance_adjust function

Adjust Estimated Variances

Adjust Estimated Variances

Calculate rescaled variances, empirical variances, etc. For use with RUV model fits.

variance_adjust(fit, ctl.idx = NULL, ebayes = TRUE, pooled=TRUE, evar = TRUE,
                rsvar = TRUE, bin = 10, rescaleconst = NULL)

Arguments

  • fit: A RUV model fit (a list), as returned by RUV2 / RUV4 / RUVinv / RUVrinv
  • ctl.idx: An index vector to specify the negative controls for use with the rescaled variances method. If unspecified, by default fit$ctl is used.
  • ebayes: A logical variable. Should empirical bayes variance estimates be calculated?
  • pooled: A logical variable. Should pooled variance estimates be calculated?
  • evar: A logical variable. Should empirical variance estimates be calculated?
  • rsvar: A logical variable. Should rescaled variance estimates be calculated?
  • bin: The bin size to use when calculating empirical variances.
  • rescaleconst: Can be used to speed up execution. See get_empirical_variances.

Returns

An RUV model fit (a list). In addition to the elements of the list returned by RUV2 / RUV4 / RUVinv / RUVrinv, the list will now contain: - sigma2.ebayes: Estimates of sigma^2 using the empirical bayes shrinkage method of Smyth (2004)

  • df.ebayes: Estimate of degrees of freedom using the empirical bayes shrinkage method of Smyth (2004)

  • sigma2.pooled: Estimate of sigma^2 pooled (averaged) over all genes

  • df.pooled: Degrees of freedom for pooled estimate

  • varbetahat: "Standard" estimate of the variance of betahat

  • varbetahat.rsvar: "Rescaled Variances" estimate of the variance of betahat

  • varbetahat.evar: "Empirical Variances" estimate of the variance of betahat

  • varbetahat.ebayes: "Empirical Bayes" estimate of the variance of betahat

  • varbetahat.rsvar.ebayes: "Rescaled Empirical Bayes" estimate of the variance of betahat

  • varbetahat.pooled: "Pooled" estimate of the variance of betahat

  • varbetahat.rsvar.pooled: "Rescaled pooled" estimate of the variance of betahat

  • varbetahat.evar.pooled: Similar to the above, but all genes used to determine the rescaling, not just control genes

  • p.rsvar: P-values, after applying the method of rescaled variances

  • p.evar: P-values, after applying the method of empirical variances

  • p.ebayes: P-values, after applying the empirical bayes method of Smyth (2004)

  • p.pooled: P-values, after pooling variances

  • p.rsvar.ebayes: P-values, after applying the empirical bayes method of Smyth (2004) and the method of rescaled variances

  • p.rsvar.pooled: P-values, after pooling variances and the method of rescaled variances

  • p.evar.pooled: Similar to the above, but all genes used to determine the rescaling, not just control genes

  • Fpvals.ebayes: F test p-values, after applying the empirical bayes method of Smyth (2004)

  • Fpvals.pooled: F test p-values, after pooling variances

  • p.BH: FDR-adjusted p-values

  • Fpvals.BH: FDR-adjusted p-values (from F test)

  • p.rsvar.BH: FDR-adjusted p-values (from p.rsvar)

  • p.evar.BH: FDR-adjusted p-values (from p.evar)

  • p.ebayes.BH: FDR-adjusted p-values (from p.ebayes)

  • p.rsvar.ebayes.BH: FDR-adjusted p-values (from p.rsvar.ebayes)

  • Fpvals.ebayes.BH: FDR-adjusted F test p-values (from Fpvals.ebayes)

  • p.pooled.BH: FDR-adjusted p-values (from p.pooled)

  • p.rsvar.pooled.BH: FDR-adjusted p-values (from p.rsvar.pooled)

  • p.evar.pooled.BH: FDR-adjusted p-values (from p.evar.pooled)

  • Fpvals.pooled.BH: FDR-adjusted F test p-values (from Fpvals.pooled)

References

Using control genes to correct for unwanted variation in microarray data. Gagnon-Bartsch and Speed, 2012. Available at: http://biostatistics.oxfordjournals.org/content/13/3/539.full.

Removing Unwanted Variation from High Dimensional Data with Negative Controls. Gagnon-Bartsch, Jacob, and Speed, 2013. Available at: http://statistics.berkeley.edu/tech-reports/820.

Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Smyth, 2004.

Author(s)

Johann Gagnon-Bartsch

Examples

## Create some simulated data
m = 50
n = 10000
nc = 1000
p = 1
k = 20
ctl = rep(FALSE, n)
ctl[1:nc] = TRUE
X = matrix(c(rep(0,floor(m/2)), rep(1,ceiling(m/2))), m, p)
beta = matrix(rnorm(p*n), p, n)
beta[,ctl] = 0
W = matrix(rnorm(m*k),m,k)
alpha = matrix(rnorm(k*n),k,n)
epsilon = matrix(rnorm(m*n),m,n)
Y = X%*%beta + W%*%alpha + epsilon

## Run RUV-inv
fit = RUVinv(Y, X, ctl)

## Get adjusted variances and p-values
fit = variance_adjust(fit)

See Also

RUV2, RUV4, RUVinv, RUVrinv, get_empirical_variances, sigmashrink

ruv packageRead PDF manual