How Model-Dependent Is the Estimate?
Triangulation check
The simulator reports what the fitted outcome model implies under a stated scenario. The triangulation check estimates that same scenario through a structurally different observational route — propensity-score subclassification built from the same measured predictors rather than the published outcome model. The two routes differ in form, but they share the sample, measured controls, and important observational limitations. When they land close, one concern narrows: the result appears less dependent on the published logit specification. Agreement is not confirmation of a causal effect, and either estimate may be too imprecise to settle the size. When they diverge, read the result cautiously; possible reasons include small samples, weak overlap, residual imbalance, or model dependence.
Loading simulator…
Couldn't load the simulator
Modeling on this page is restricted to binary (yes/no) outcomes.
Auto-Build Actionable found no movable predictors for this outcome, so this simulator uses the best Standard model instead — its predictors may not be classified as directly movable.
Set the scenario
The number on each card is the variable's incremental score as a share of its standalone score — how much of its one-at-a-time separation on the outcome remains once the other variables are included. Bars and ordering use the incremental score itself; the share is a display convention. Cards are ordered largest incremental first.
Predicted outcome
How did you change the survey responses?
Each pair of bars compares the actual distribution (green) to the distribution under your scenario (red where you changed it). This is what you changed — not the predicted impact, just the input.
How certain is this result?
Every prediction has wiggle room — these histograms show how much. The green bars are the plausible answers for the baseline; the red bars are the plausible answers for your scenario. Where the colors overlap, the two answers are close enough that the model can't cleanly tell them apart.
Set-everyone-to-X table
Click to show predicted outcomes for every level of every variable
Set-everyone-to-X table
Click to show predicted outcomes for every level of every variable
For each variable's level, the predicted outcome if every respondent had that response, all else unchanged. This is the analytic view of the simulator.
Calibration: predicted vs. observed
How closely the model's predicted probabilities track the observed outcome rates, binned by predicted decile. Bubble size shows respondents per bin. The dashed diagonal is perfect calibration; the blue scenario line is your run; the gray line (when present) is the unmodified base model for comparison.
Explore each factor
For each variable, see how the predicted outcome would change if you switched just that one selection — holding all your other selections fixed. This shows which individual factor settings produce the largest model-implied changes in the current scenario.
Separate from the scenario cross-check below. Pick one factor, and this compares the people who answered one way with otherwise-similar people who answered another — using the model’s remaining factors to line them up.
Pick things that were already true about people before this answer — age, background, other long-held views. Don't pick something that could be downstream of this answer; adjusting for it could distort the comparison.
These are the model’s other predictors. Both methods adjust for them so the comparison uses the same measured controls.
Run a scenario above and we’ll cross-check that exact change here — the simulator’s number beside an independent propensity estimate.