Description Usage Arguments Details Value Author(s) See Also Examples

View source: R/correlateNull.R

Build a distribution of correlations under the null hypothesis of independent expression between pairs of genes.

1 2 3 4 5 6 7 8 | ```
correlateNull(
ncells,
iters = 1e+06,
block = NULL,
design = NULL,
equiweight = TRUE,
BPPARAM = SerialParam()
)
``` |

`ncells` |
An integer scalar indicating the number of cells in the data set. |

`iters` |
An integer scalar specifying the number of values in the null distribution. |

`block` |
A factor specifying the blocking level for each cell. |

`design` |
A numeric design matrix containing uninteresting factors to be ignored. |

`equiweight` |
A logical scalar indicating whether statistics from each block should be given equal weight.
Otherwise, each block is weighted according to its number of cells.
Only used if |

`BPPARAM` |
A BiocParallelParam object that specifies the manner of parallel processing to use. |

The `correlateNull`

function constructs an empirical null distribution for Spearman's rank correlation when it is computed with `ncells`

cells.
This is done by shuffling the ranks, calculating the correlation and repeating until `iters`

values are obtained.
No consideration is given to tied ranks, which has implications for the accuracy of p-values in `correlatePairs`

.

If `block`

is specified, a null correlation is created within each level of `block`

using the shuffled ranks.
The final correlation is then defined as the average of the per-level correlations,
weighted by the number of cells in that level if `equiweight=FALSE`

.
Levels with fewer than 3 cells are ignored, and if no level has 3 or more cells, all returned correlations will be `NA`

.

If `design`

is specified, the same process is performed on ranks derived from simulated residuals computed by fitting the linear model to a vector of normally distributed values.
If there are not at least 3 residual d.f., all returned correlations will be `NA`

.
The `design`

argument cannot be used at the same time as `block`

.

A numeric vector of length `iters`

is returned containing the sorted correlations under the null hypothesis of no correlations.

Aaron Lun

`correlatePairs`

, where the null distribution is used to compute p-values.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 | ```
set.seed(0)
ncells <- 100
# Simplest case:
null.dist <- correlateNull(ncells, iters=10000)
hist(null.dist)
# With a blocking factor:
block <- sample(LETTERS[1:3], ncells, replace=TRUE)
null.dist <- correlateNull(block=block, iters=10000)
hist(null.dist)
# With a design matrix.
cov <- runif(ncells)
X <- model.matrix(~cov)
null.dist <- correlateNull(design=X, iters=10000)
hist(null.dist)
``` |

Embedding an R snippet on your website

Add the following code to your website.

For more information on customizing the embed code, read Embedding Snippets.