Skip to contents

Bootstrap an HTNA Network

Source: R/bootstrap_htna.R
bootstrap_htna.Rd

Thin wrapper around Nestimate::bootstrap_network() that runs the standard state-level edge bootstrap on an htna network and tags the result so it is identifiable as an htna bootstrap. The bootstrap inference itself is Nestimate's: per-edge mean, sd, p-value, significance, and confidence / credibility intervals across iter resamples. State-level granularity is the correct level for edge-stability analysis - aggregating up to actor pairs would smear over real edge-by-edge differences.

Usage

bootstrap_htna(x, ...)

# S3 method for class 'htna'
bootstrap(x, ...)

Arguments

x

[htna]
A network built with build_htna().

...

Arguments forwarded to Nestimate::bootstrap_network() (e.g. iter, ci_level, consistency_range, seed).

Value

An object of class c("htna_bootstrap", "net_bootstrap"). All slots produced by Nestimate::bootstrap_network() are preserved verbatim ($summary, $mean, $sd, $p_values, $significant, $ci_lower/$ci_upper, $cr_lower/$cr_upper, $model, $original, etc.).

Details

What this wrapper adds is identity and downstream class continuity:

  • The returned object has class "htna_bootstrap" ahead of "net_bootstrap", so inherits(boot, "htna_bootstrap") works.

  • The actor partition ($nodes$groups, $node_groups) is restored on boot$model from the input network so cograph's auto-detect still recognizes the heterogeneous schema downstream.

  • boot$model is re-tagged with "htna" at the front of its class chain so the chain is c("htna", "netobject", "cograph_network") - identical to what build_htna() returns.

See also

Nestimate::bootstrap_network() for the underlying algorithm and slot documentation. build_htna() for constructing the input.

Examples

# \donttest{
data(human_ai)
net  <- build_htna(human_ai, actor_type = "actor_type")
#> Warning: A network with one long sequence is not recommended and can't be validated using bootstrap and other confirmatory testings.
#> Metadata aggregated per session: ties resolved by first occurrence in 'session_date' (1 sessions), 'cluster' (42 sessions), 'actor_type' (24 sessions)
boot <- bootstrap_htna(net, iter = 50)
inherits(boot, "htna_bootstrap")            # TRUE
#> [1] TRUE
inherits(boot$model, "htna")                # TRUE
#> [1] TRUE
head(boot$summary)                          # state-level edge stability
#>   from        to     weight       mean          sd    p_value   sig   ci_lower
#> 1  Ask       Ask 0.01823579 0.01802025 0.003137153 0.13725490 FALSE 0.01276864
#> 2  Ask     Check 0.06318914 0.06327862 0.005019691 0.01960784  TRUE 0.05248738
#> 3  Ask   Execute 0.02205259 0.02182131 0.003134294 0.11764706 FALSE 0.01523540
#> 4  Ask Frustrate 0.11577608 0.11634468 0.007946139 0.01960784  TRUE 0.09850669
#> 5  Ask   Inquire 0.06318914 0.06201186 0.005187685 0.01960784  TRUE 0.05072386
#> 6  Ask      Plan 0.40882103 0.41117163 0.015487179 0.01960784  TRUE 0.38479470
#>     ci_upper   cr_lower   cr_upper
#> 1 0.02308084 0.01367684 0.02279474
#> 2 0.07297633 0.04739186 0.07898643
#> 3 0.02760946 0.01653944 0.02756573
#> 4 0.13183281 0.08683206 0.14472010
#> 5 0.07167155 0.04739186 0.07898643
#> 6 0.43791033 0.30661578 0.51102629
# }