Wald Likelihood Support

Explore how much relative support an observed estimate and reported 95% confidence interval provide for candidate effect sizes under a one-parameter Wald reconstruction.

This educational research tool shows normalized, approximate relative likelihood. It does not recover the original fitted-model likelihood, assign probabilities of truth, validate clinical thresholds, or provide clinical decision support.

Observed estimate and interval

Supply a published two-sided 95% confidence interval. The point estimate is optional because the interval midpoint determines the reconstructed estimate on the working scale.

Effect measure

Point estimate (optional)

Lower 95% CI

Upper 95% CI

Null value (optional)

Defaults to 1 for ratios and 0 for additive effects.

Reference thresholds (optional)

Separate values with commas or spaces. These are user-supplied reference markers, not validated clinical or decision thresholds.

Support-interval criterion

S−2 uses a log-support cutoff of −2. A 2:1 interval is a different, narrower criterion.

Optional candidate A versus candidate B

Supply both candidates. Results are reported explicitly in the order L(A)/L(B), with the finite log ratio retained if the ordinary ratio cannot be represented.

Candidate A

Candidate B

Display lower (optional)

Display upper (optional)

A plausible display range changes only the plotted and exported grid, never the reconstructed support summaries or interval.

Advanced display controls

Ratio-axis spacing

Curve view

Grid points

Relative-support result

Reconstructed summary

Selected support interval

The interval contains candidate values for which the CI-implied estimate is no more than R times as supported under the reconstructed Wald likelihood.

Null and reference support

Normalized relative support at the null and user reference values
Reference / display value	Working value	Relative likelihood vs estimate	Log relative likelihood	Estimate / candidate	Candidate / null	From estimate	From null	Inside selected interval

Candidate A versus candidate B

No pairwise comparison supplied. Add both candidate A and candidate B to report log L(A)/L(B) and L(A)/L(B).

Pairwise relative support in the explicit order L(A)/L(B)
Candidate A	Candidate B	log L(A)/L(B)	L(A)/L(B)	Direction

Figure caption

Exports

CSV contains only the five plotted grid fields. Interval and pairwise results remain in the accessible tables and strict JSON response.

When the local runtime is ready, update the support curve to populate the plot, reconstructed summary, support tables, caption, and exports.

Relative likelihood is normalized to 1 at the CI-implied estimate. It is not a posterior probability that a candidate value is true.