Total Variance Model

v20a.28
Raw source variable=— · T=— · Runs=—
Network family N=— · Networks=—
Display

1. Evidence source X

Purpose. Generate the reusable world/run-level evidence source X before any pooling, thresholding, or credence regime is applied.

Active setup

Inputs. Choose the raw random variable, testing power, and run count in the setup modal. The raw panel only displays the active setup and generated results.
—
raw mean
0.535000
EV = Var(X)
0.248775
T
240
Var( X̄ ) pred
0.001037
E[ X̄ ] emp
—
Var( X̄ ) emp
—
single-trial EV = Var(X), while the raw run mean has predicted variance EV/TOTAL
Generate the raw world evidence first. Population and network structure only enter when downstream regimes partition that evidence.

Raw source view

This is the pre-pooling X source. Use the x-axis modes to switch between world means and the underlying raw studies before any downstream agent partitioning.

2. Evidence regime: X

Purpose. Apply the selected network family to X and decompose EV(X) into IV + SV + MV.

Network selection

Inputs. Network population and family selection live in the network modal. Changing population repartitions the generated world evidence for downstream pooling.
slice s=0
(k,s)
—
m
1
Δ = EV − Σ
—
EV(X) pred / emp
—
EV(X) = IV + SV(X) + MV(X)
For s>0, the X regime still uses analytic overlap-based predictions. Later-regime predictions are preserved where available and otherwise fall back to experimental verification.

X-regime variance decomposition

View. Light bars are analytic predictions. Dark bars are simulated verification for the same network.
show

3. Judgment regime: Jθ

Purpose. Threshold the network-shaped evidence into Jθ = 1(X > θ). Compare EV(Jθ) = SV(Jθ) + MV(Jθ) against the EV(X) baseline.

Jθ-regime variance decomposition

View. Filled bars are Jθ results. Baseline markers show EV(X) for comparison when enabled.
show

4. Credence regime: Cθ

Purpose. Represent threshold judgments as credences Cθ. Compare EV(Cθ) = SV(Cθ) + MV(Cθ) against the EV(X) baseline.

Prior P(Jθ=1)

Inputs. Choose the starting prior and whether agents share it or initialize independently around it.
initialize
E[Cθ] emp
—
EV(Jθ)
—
EV(Cθ)
—
SV(Cθ) + MV(Cθ) = EV(Cθ), starting from the prior above

Cθ-regime variance decomposition

View. Filled bars are Cθ results. EV(X) baseline shows the X-regime comparison budget.
show

5. Beta credence regime: Cβθ

Purpose. Randomize the starting prior with a Beta distribution, then represent the same threshold judgments as Cβθ.

Beta prior

Inputs. α and β set the prior shape. Together uses one shared draw per run; separate draws one prior per agent.
prior draw
E[Cβθ] emp
—
Var(Cβθ)
—
Var(Cβθ) = SV(Cβθ) + MV(Cβθ)

Cβθ-regime variance decomposition

View. Filled bars are Cβθ results. EV(X) baseline shows the X-regime comparison budget.
show
Single parameter sweep
θ comes from the active random-variable threshold; Cθ and Cβθ prior controls open here.
vary this parameter
Variance variable
Bernoulli varies ε; Normal varies σ².
vary this parameter
Testing power
variable=— · T=— · Runs=—
vary this parameter
Network size
N=— · network=—
vary this parameter
Connectivity variable
fixed (k=0, s=0)
Display

1. Single parameter sweep source

Configure the four sweep parameters in the top boxes. Select exactly one to vary; the others are fixed for this single-parameter sweep.

Active setup

Inputs live in the four top boxes. This panel reports the active random variable and generated values.
The active single-sweep setup is controlled in the top boxes.
raw mean
—
raw EV
—
T values
—
Within each later figure, the x-axis is the selected varied parameter. The raw source remains world-level; downstream regimes partition or regenerate evidence according to the active sweep variable.

Raw source view

Default view shows world/run means. Switch the x-axis mode to inspect the underlying raw evidence itself; when a sweep point is highlighted, the visible raw evidence is emphasized and the rest fades back where a shared source is meaningful.
Choose one parameter box to vary, then run the single-parameter sweep.

2. Evidence regime: X

X regime report

Evidence after network sharing. This is the baseline regime for downstream comparisons.
regime
EV(X)
baseline
none
components
IV(X) + SV(X) + MV(X)
network
(k=0, s=0)

X-regime single-sweep profile

3. Judgment regime: Jθ

Jθ regime report

Threshold judgment derived from X. The EV(X) baseline is a comparison budget, not an upstream regime.
regime
EV(Jθ)
baseline
EV(X)
components
SV(Jθ) + MV(Jθ)
θ
—
network
—

Jθ regime single-sweep profile

4. Credence regime: Cθ

Cθ regime report

Controls for the judgment prior live in the top-bar downstream-regime modal. This panel reports the generated Cθ decomposition.
regime
EV(Cθ)
baseline
EV(X)
components
SV(Cθ) + MV(Cθ)
θ
—
Cθ prior
—
network
—

Cθ regime single-sweep profile

5. Beta credence regime: Cβθ

Cβθ regime report

Controls for the beta prior live in the top-bar downstream-regime modal. This panel reports the generated Cβθ decomposition.
regime
EV(Cβθ)
baseline
EV(X)
components
SV(Cβθ) + MV(Cβθ)
θ
—
Cβθ prior
—
network
—

Cβθ regime single-sweep profile

Two parameter sweep summary
fixed—
grid x—
grid y—
within cell—
Display
Compute

Two parameter sweep

grid y-axis
N
grid x-axis
connectivity
within-cell x-axis
samples T
fixed parameter
—
runs / point
—
row values —
column values —
x-axis values —
Open Edit params, choose the two-parameter sweep roles, and then run the two parameter sweep.

X-regime two-parameter sweep

Jθ-regime two-parameter sweep

Grid setup
auto: N=4, 12, 36 · k+s≤7
Parameters implied by sweep choice
Display

Network grid sweep

Each cell is a network in the k/s diamond. Choose one variable to sweep inside every cell; the cell can show the network itself, the sweep result, or the result over a faded network diagram.

Choose the within-cell sweep variable; the tab will compute N=4, 12, and 36 automatically, capped at k+s≤7.