Illustrates how network properties vary with the similarity threshold

make_threshold_profile() outputs properties of the agent or symbolic network as a function of similarity threshold.

Usage

make_threshold_profile(
  data,
  layer = NULL,
  comparisons = NULL,
  metric = NULL,
  count = NULL,
  limits = NULL,
  dummycode = NULL,
  ...
)

Arguments

data

A data frame corresponding to the attitudes held by agents with respect to a number of items

layer

A string flag specifying the type of network to be extracted,

• "agent" produces the network corresponding to the agents, which we assume to be rows in data

• "symbolic" produces the network corresponding to the symbols, or items, which we assume to be columns in data

comparisons

An integer, minimum number of comparisons for valid distance.

metric

A string option describing the similarity metric to be used.

count

The number of threshold values to include in the description.

limits

Specify the limits of the Likert range in during a data preprocessing step.

dummycode

Specify whether to apply dummycoding during a data preprocessing step.

...

Used to handle alternative argument spellings.

Value

A data frame containing properties of the agent or symbolic network as a function of the similarity threshold. In particular, it contains three columns named

• threshold, the value of the similarity threshold

• ad, the average degree resulting from threshold, and

• lcc, the size of the largest connected component resulting from threshold

Details

Note that this routine is expensive on large graphs. We study networks over the full range of similarity thresholds [-1, 1], and as a result, produce networks that are complete at the lower limit of that range. Note that by default we will subsample the provided survey with the C++ implementation in order to avoid memory issues. We could then allow a flag that turns off the subsampling step, at the user's peril.

Examples

S <- make_synthetic_data(20, 5)