Quantitative Research · Portfolio Project
Latent Liquidity &
Crisis Dashboard
A Bayesian state-space model that estimates systemic liquidity stress and forward crisis probability from market microstructure data. Latent factors h and z are inferred via MCMC; crisis signals are generated as 60-day predictive probabilities.
Bayesian State-Space
MCMC Inference
Latent Factor Model
Crisis Prediction
Peak Crisis Probability
—
loading…
Max Liquidity Stress Lt
—
loading…
Avg Tail Probability P(L>2)
—
across selected window
Data Window
—
observations in range
Filter Time Window
Select a range or use an era preset
Crisis Probability (next 60 days)
Dashed line = 0.50 thresholdRendering…
Liquidity Stress Lt with 90% CI
Green dashed line = threshold L=2Rendering…
Latent Factors h, z
Posterior mean estimatesRendering…
Pre-rendered Outputs
Click any image to enlarge
experiments/visualize_crisis.py
Reproduce Results
$
python3 main.py predict
# recompute crisis data → results/crisis_res.npz
$
python3 experiments/visualize_crisis.py
# regenerate static plots
$
python3 experiments/animate_crisis.py
# create animated GIFs
$
python3 scripts/export_json.py
# export crisis_res.json for this dashboard