.prior_kmeans: Elicits data-driven priors from a flowSet object for specified channels using the K-Means clustering algorithm

Description

We elicit data-driven prior parameters from a flowSet object for specified channels. For each sample in the flowSet object, we apply kmeans to obtain K clusters. From each cluster, we determine its centroid and the sample covariance matrix. We then aggregate these two sample moments across all samples for each cluster.

Usage

.prior_kmeans(flow_set, channels, K, nu0 = 4, w0 = 10, nstart = 10,
  pct = 0.1, min = NULL, max = NULL, ...)

Arguments

flow_set

a flowSet object

channels

a character vector containing the channels in the flowSet from which we elicit the prior parameters for the Student's t mixture

the number of mixture components to identify

nu0

prior degrees of freedom of the Student's t mixture components.

the number of prior pseudocounts of the Student's t mixture components.

nstart

number of random starts used by kmeans algorithm

pct

percentage of randomly selected cells in each flowFrame that is used to elicit the prior parameters. The value should must be greater than 0 and less than or equal to 1.

min

a numeric vector that sets the lower bounds for data filtering. If NULL (default), no truncation is applied.

max

a numeric vector that sets the upper bounds for data filtering. If NULL (default), no truncation is applied.

...

Additional arguments passed to kmeans

Value

list of flowClust prior parameters

Details

Because the cluster labels returned from kmeans are arbitrary, we align the clusters based on the centroids that are closest to a randomly selected reference sample. We apply the Hungarian algorithm implemented using the solve_LSAP function from the clue package to assist with the alignment.

If each frame within flow_set has a large number of cells, the computational costs of kmeans can be a burden. We provide the option to randomly select pct, a percentage of the cells from each flow frame to which kmeans is applied.