Package 'INTRIGUE' reference manual

Title:	Quantify and Control Reproducibility in High-Throughput Experiments
Description:	Estimate the proportions of the null and the reproducibility and non-reproducibility of the signal group for the input data set. The Bayes factor calculation and EM (Expectation Maximization) algorithm procedures are also included.
Authors:	Yi Zhao [aut], Xiaoquan Wen [aut], Michael Kleinsasser [cre]
Maintainer:	Michael Kleinsasser <[email protected]>
License:	GPL-2
Version:	0.1.0
Built:	2025-02-05 02:54:27 UTC
Source:	https://github.com/cran/INTRIGUE

Bayes Factor Approximation

Description

A function calculates the approximation for bayes factor, when the value of original bayes factor goes to infinity.

Usage

bf.approx(z, param, size, k2, oa2)
bf.approx(z, param, size, k2, oa2)

Arguments

`z`	The index for individual(i).
`param`	Input dataset.
`size`	Number of replicates(m).
`k2`	Grid value of $k^2$ .
`oa2`	Grid value of $\omega^2$ .

Value

Approximation for bayes factor in log scale.

Bayes Factor Calculation Scheme for CENF prior

Description

A function that calculates bayes factor for each data pair on each grid point in log scale.

Usage

bf.cal.cefn(data, hyperparam)
bf.cal.cefn(data, hyperparam)

Arguments

`data`	A dataset which is constructed by pairs of coefficient values $\beta$ and standard errors $se(\beta)$ .
`hyperparam`	A two-dimensional vector denoting all the grid points, namely, $k$ x $\omega$ .

Value

A list records all the log scale bayes factor values.

Bayes Factor Calculation Scheme for META prior

Description

A function that calculates bayes factor for each data pair on each grid point in log scale.

Usage

bf.cal.meta(data, hyperparam = NULL, bf.only = FALSE)
bf.cal.meta(data, hyperparam = NULL, bf.only = FALSE)

Arguments

`data`	A dataset which is constructed by pairs of coefficient values $\beta$ and standard errors $se(\beta)$ .
`hyperparam`	A two-dimensional vector denoting all the grid points, namely, $\phi$ x $\omega$ .
`bf.only`	A boolean, denoting whether this function is called to calculate Bayes factor for META prior only. Usually used when publication bias issue is the target.

Value

A list records all the log scale bayes factor values or a list records log scale bayes factor for null, reproducible and irreproducible model (when bf.only=TRUE).

Bayes Factor EM Updating Scheme

Description

A function that describes the updating process in E step and M step for EM algorithm. It will be used in SQUAREM package.

Usage

bf.em(w, bf)
bf.em(w, bf)

Arguments

`w`	The weight vector in previous M step.
`bf`	A vector recording all the bayes factor values in log scale.

Value

The updated weight vector in current M step(wnew).

Bayes Factor Loglikelihood Function

Description

Calculate the updated loglikelihood value in EM algorithm, and to evaluate whether converge or not.

Usage

bf.loglik(w, bf)
bf.loglik(w, bf)

Arguments

`w`	The current weight vector
`bf`	A vector recording all the bayes factor values in log scale.

Value

Negative summation of loglikelihood values.

Bayes Factor Weighted Summation

Description

A function calculates the weighted summation of bayes factor.

Usage

bf.weighted_sum(w, bf, i)
bf.weighted_sum(w, bf, i)

Arguments

`w`	Input weight vector.
`bf`	Input bayes factor vector
`i`	Individual index.

Value

Weighted sum for bayes factor in log scale.

Heterogeneity Evaluation

Description

Evaluating the overall and individually heterogeneity and reproducibility for the given individuals(units) shared in different replicates.

Usage

hetero(
  data,
  use_cefn = TRUE,
  rep = NULL,
  irre = NULL,
  phi_min = NULL,
  phi_max = NULL,
  sq_em_tol = 1e-04,
  fdr.level = NULL,
  sample_size = NULL
)
hetero(
  data,
  use_cefn = TRUE,
  rep = NULL,
  irre = NULL,
  phi_min = NULL,
  phi_max = NULL,
  sq_em_tol = 1e-04,
  fdr.level = NULL,
  sample_size = NULL
)

Arguments

`data`	A dataset which is constructed by pairs of coefficient values $\beta$ and standard errors $se(\beta)$ .
`use_cefn`	A boolean, denoting whether to use CEFN prior. If the value is TRUE, CEFN prior is used, else, META prior is applied. The default value is TRUE.
`rep`	A vector, denoting all the $k^2$ (under CEFN prior) or $r$ (under META prior) values constructing the reproducible signals. If not specified, the default one is c(0.105,0.260,0.369), which corresponds to the several prior values satisfy that $Pr(\beta_{i,1}, \beta_{i,2} have a same sign)=0.99, 0.975, 0.95$ for CEFN prior.
`irre`	A vector, denoting all the $k^2$ or $r$ values constructing the irreproducible signals. If not specified, the default one is c(2.198, 3.636, 6.735), which corresponds to the several prior values satisfy that $Pr(\beta_{i,1}, \beta_{i,2} have a same sign)=0.75, 0.70, 0.65$ for CEFN prior.
`phi_min`	A value which determines the maximum $phi$ . If not specified, will be constructed from the input datasets.
`phi_max`	A value which determines the minimum $phi$ . If not specified, will be constructed from the input datasets.
`sq_em_tol`	A small, positive scalar that determines when iterations should be terminated in squarem algorithm. The default value is $1e-4$ .
`fdr.level`	The user-defined rejection level for false discovery rate.
`sample_size`	The user-defined sample size.

Value

A list with the following components:

`gridweight`	The final optimal weight vector evaluated on each grid point.
`ind_prob`	A matrix denoting the converged probability for each individual being inside the three different groups, namely, the null, the reproducible and the irreproducible group.
`est_prop`	The estimated proportion value for the three different groups, namely, the null, the reproducible and the irreproducible group.
`lfdr`	The local false discovery rate based on the null hyppthesis of unit belonging to $H_R$ , reproducible group. $lfdr=1-Pr(H_R)$
`significant`	If fdr.level is specified, a significant object recording True or False will be returned

Examples

data("heterodata")

hetero.out<-hetero(heterodata,fdr.level=0.05)
names(hetero.out)
print(hetero.out$est_prop)


## for CRAN check
hetero.out<-hetero(heterodata[1:100,],fdr.level=0.05)


data("heterodata")

hetero.out<-hetero(heterodata,fdr.level=0.05)
names(hetero.out)
print(hetero.out$est_prop)


## for CRAN check
hetero.out<-hetero(heterodata[1:100,],fdr.level=0.05)

Local False Discovery Rate Evaluation

Description

Local False Discovery Rate Evaluation

Usage

hetero.lfdr(cat, fdr.level)
hetero.lfdr(cat, fdr.level)

Arguments

`cat`	The final individual-level probabilities of falling into three categories, separately.
`fdr.level`	Rejection level for Local false discovery rate, if not specified, the rejection decision procedure won't be run.

Value

A list that preserves local false discovery rate and 5tthe corresponding reject decision if called.

Simulation Dataset

Description

This is a simulation dataset, containing n=5000 units and m=2 replicates. The true proportion for the null, the reproducible and the irreproducible group is 0.80, 0.18, 0.02 separately.

Usage

data("heterodata")
data("heterodata")

Format

An object of class matrix (inherits from array) with 1000 rows and 4 columns.

Examples

data("heterodata")
data("heterodata")

Package 'INTRIGUE'

Help Index

Bayes Factor Approximation

Description

Usage

Arguments

Value

Bayes Factor Calculation Scheme for CENF prior

Description

Usage

Arguments

Value

Bayes Factor Calculation Scheme for META prior

Description

Usage

Arguments

Value

Bayes Factor EM Updating Scheme

Description

Usage

Arguments

Value

Bayes Factor Loglikelihood Function

Description

Usage

Arguments

Value

Bayes Factor Weighted Summation

Description

Usage

Arguments

Value

Heterogeneity Evaluation

Description

Usage

Arguments

Value

Examples

Local False Discovery Rate Evaluation

Description

Usage

Arguments

Value

Simulation Dataset

Description

Usage

Format

Examples