tsum_test: Generate T sum statistics and p-values from simulation.

In bahlolab/exSTRa: Expanded STR algorithm: detecting expansions in Illumina sequencing data

Generate T sum statistics and p-values from simulation.

Description

When applied to an exstra_score object, T sum statistics are calculated as described in Tankard et al. May also be applied on a pre-existing exstra_tsum that will regenerate the values.

Usage

tsum_test(
  strscore,
  trim = ifelse(case_control, trim.cc, trim.all),
  trim.all = 0.15,
  trim.cc = 0,
  min.quant = 0.5,
  give.pvalue = TRUE,
  B = 999,
  correction = c("bf", "loci", "samples", "uncorrected"),
  alpha = 0.05,
  case_control = FALSE,
  early_stop = TRUE,
  early_A = 0.25,
  early_stop_min = 50,
  parallel = FALSE,
  cluster_n = NULL,
  cluster = NULL,
  keep.sim.tsum = FALSE
)

Arguments

strscore	An exstra_score object.
trim	Trim this proportion of data points at each quantile level (rounded up). Must be at least 0 and less than 0.5, but values close to 0.5 may remove all samples and hence result in an error.
trim.all	Trim value when looking for outliers in all samples. Default 0.15.
trim.cc	Trim value used in case-control analysis. Default of 0.
min.quant	Only quantiles above this value are used in constructing the statistic.
give.pvalue	Whether to calculate the p-value. As this can be slow it can be useful to turn off if only the t sum statistics are required.
B	Number of simulations in calculating null distributions. The denominator will be B + 1, hence values of B = 10^i - 1 will result in p-values that are decimal fractions.
correction	Correction method of p_value() function.
alpha	Signficance level of p_value() function.
case_control	If TRUE, only calculate for samples designated cases. Otherwise all samples are used to calculate the background distribution.
early_stop	Simulation may use less replicates when all p-values are large, controlled with early_A.
early_A	Simulations may stop when p.value.sd < early_A * min(p.value). Checked approximately when the number of simulations has doubled.
early_stop_min	Minimum number of simulations to run before early termination.
parallel	Use the parallel package when simulating the distribution, creating the required cluster. If cluster is specified then this option makes no difference.
cluster_n	If parallel is TRUE, then the number of nodes in the cluster is automatically set as half of those available on your machine (but never less than 1). This option allows manual setting of the number of nodes, either less to free up other resources, or to maximize available resources. If cluster is specified then this option makes no difference.
cluster	A cluster object from the parallel package. Use if you wish to set up the cluster yourself or reuse an existing cluster.
keep.sim.tsum	For inspection of simulations. If TRUE, keep all simulation Tsum statistics in output$xecs (default FALSE).

Value

An exstra_tsum object with T statistics and p-values (if calculated).

References

Rick M. Tankard, Martin B. Delatycki, Paul J. Lockhart, Melanie Bahlo. Detecting known repeat expansions with standard protocol next generation sequencing, towards developing a single screening test for neurological repeat expansion disorders. bioRxiv 157792; doi: https://doi.org/10.1101/157792

See Also

tsum_p_value_summary

Examples

exp_test <- tsum_test(exstra_wgs_pcr_2[c("HD", "SCA6")], B = 50)
exp_test

## Not run: 
exp_test_parallel <- tsum_test(exstra_wgs_pcr_2[c("HD", "SCA6")], parallel = TRUE, B = 999)
exp_test_parallel

## End(Not run)

bahlolab/exSTRa documentation built on Sept. 17, 2022, 5:08 p.m.