Replace 5 inline ESS calculations with calc_model_ess()

[gilesjohnr]

Context

The MOSAIC package has a dedicated calc_model_ess() function that computes Effective Sample Size using Kish or perplexity methods. During an audit (similar to the 16 inline cor()^2 calls that were replaced with calc_model_R2() in v0.22.4), we found 5 inline ESS calculations that should use the utility function.

Problem

These files compute ESS inline as 1 / sum(weights^2) instead of calling calc_model_ess():

1. R/grid_search_best_subset.R — lines 157, 221
2. R/plot_model_parameter_sensitivity.R — line 99
3. R/plot_model_posteriors_detail.R — lines 346, 356

(R/calc_bfrs_posterior.R:93 also has one, but that file is orphaned — see issue #76)

Required Actions

1. Replace each inline 1 / sum(weights^2) with calc_model_ess(weights, method = "kish")
2. Verify the utility function handles edge cases (zero weights, single weight) that the inline code may not
3. Run devtools::test() to verify
4. Grep for any other inline ESS patterns: 1/sum.*\\^2, sum.*weights.*\\^2

Error Classification

AI-introduced error of omission — same pattern as the calc_model_R2() issue: utility function exists but inline duplicates persist.

🤖 Generated with Claude Code