intermediate_stability: intermediate_stability
In microbiome/microbiomeold: Tools for microbiome analysis

Description Usage Arguments Details Value Author(s) Examples

Quantify stability with respect to a given reference point for all variables.

1	intermediate_stability(dat, meta, reference.point = NULL, method = "lm")

`dat`	Input data matrix (variables x samples)
`meta`	Metadata (samples x factors). This should contain for each sample the following self-explanatory fields: subjectID and time.
`reference.point`	Calculate stability of the data w.r.t. this point. By default the intermediate range is used (min + (max - min)/2)
`method`	"lm" (linear model) or "correlation"; the linear model takes time into account as a covariate

This method decomposes the data set into differences between consecutive time points. For each variable and time point we calculate for the data values: (i) the distance from reference point; (ii) distance from the data value at the consecutive time point. The "correlation" method calculates correlation between these two variables. Negative correlations indicate that values closer to reference point tend to have larger shifts in the consecutive time point. The "lm" method takes the time lag between the consecutive time points into account as this may affect the comparison and is not taken into account by the straightforward correlation. Here the coefficients of the following linear model are used to assess stability: abs(change) ~ time + abs(start.reference.distance)

A list with following elements: stability: vector listing the intermediate stability for each variable data: processed data sets used for calculations

Leo Lahti leo.lahti@iki.fi

# Create simulated example data
  dat <- matrix(rnorm(1000), nrow = 10)
  rownames(dat) <- paste("Variable", 1:nrow(dat), sep = "")
  colnames(dat) <- paste("Sample", 1:ncol(dat), sep = "")
  meta <- data.frame(list(
    	  sampleID = colnames(dat),
  	  subjectID = rep(paste("subject", 1:50, sep = "-"), each = 2),
	  time = rep(1:2, 50)))
  # Intentionally make point very unstable around 0
  dat[1, meta$time == 2] <- 1/abs(dat[1, meta$time == 1])
  s <- intermediate_stability(dat, meta, reference.point = 0)