egutierrez
47fac22230
- .claude/CLAUDE.md - .claude/commands/subagentes.md - .claude/rules/INDEX.md - .mcp.json - bash/functions/cybersecurity/analyze_dns.md - bash/functions/cybersecurity/audit_http_headers.md - bash/functions/cybersecurity/audit_ssh_config.md - bash/functions/cybersecurity/check_firewall.md - bash/functions/cybersecurity/detect_suspicious_users.md - bash/functions/cybersecurity/encrypt_file.md - ... Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
70 lines
3.2 KiB
Markdown
70 lines
3.2 KiB
Markdown
---
|
|
name: stats_summary
|
|
kind: function
|
|
lang: cpp
|
|
domain: datascience
|
|
version: "1.0.0"
|
|
purity: pure
|
|
signature: "double stats_sum(const double*, size_t); double stats_mean(const double*, size_t); double stats_min(const double*, size_t); double stats_max(const double*, size_t); double stats_variance(const double*, size_t, bool sample=true); double stats_std(const double*, size_t, bool sample=true); double stats_quantile(const double*, size_t, double p); double stats_quantile_sorted(const double*, size_t, double p); double stats_percentile(const double*, size_t, double pct); void stats_sort(const double*, size_t, double* out)"
|
|
description: "Estadistica descriptiva pura sobre arrays double: sum (Kahan), mean, min, max, variance/std (Welford one-pass, sample/poblacional), quantile (R type-7) y percentile. stats_sort externalizable para evitar copias en queries multiples."
|
|
tags: [stats, mean, variance, std, quantile, percentile, welford, datascience, pendiente-usar]
|
|
uses_functions: []
|
|
uses_types: []
|
|
returns: []
|
|
returns_optional: false
|
|
error_type: ""
|
|
imports: [cstddef, cmath, algorithm, vector, cstring]
|
|
tested: false
|
|
tests: []
|
|
test_file_path: ""
|
|
file_path: "cpp/functions/datascience/stats_summary.cpp"
|
|
params:
|
|
- name: data
|
|
desc: "Array de doubles (no necesariamente ordenado salvo *_sorted)."
|
|
- name: n
|
|
desc: "Tamano del array. n=0 devuelve identidades sensatas (0 para sum/mean/min/max/var/std)."
|
|
- name: sample
|
|
desc: "Variance/std: true = muestral (n-1), false = poblacional (n). Default true."
|
|
- name: p
|
|
desc: "Quantile en [0, 1]. Valores fuera se clampean."
|
|
- name: pct
|
|
desc: "Percentile en [0, 100]. Internamente p = pct/100."
|
|
- name: out
|
|
desc: "(stats_sort) buffer destino. Si out == data, ordena in-place."
|
|
output: "Escalar (double) con la estadistica solicitada. stats_sort modifica out in-place; el resto no muta data."
|
|
---
|
|
|
|
# stats_summary
|
|
|
|
Pack de estadisticas basicas sobre arrays raw. Diseñado para post-proceso de samples MC, sesiones de simulacion, cadenas MCMC.
|
|
|
|
## Performance
|
|
|
|
- `stats_sum`: Kahan summation (O(n), ~5% mas lento que sum naive pero sin drift en sumas de millones de fp64).
|
|
- `stats_variance`: Welford one-pass (O(n), una sola pasada). No hay las cancelaciones catastroficas del E[X^2] - E[X]^2 naive.
|
|
- `stats_quantile`: O(n log n) por copia + sort. Para multiples queries del mismo dataset, llamar `stats_sort` una vez y `stats_quantile_sorted` despues — O(n log n + Q).
|
|
|
|
## Patron tipico
|
|
|
|
Resumen de un session simulator (vr_tiered_lab):
|
|
|
|
```cpp
|
|
std::vector<double> pnls(N);
|
|
// ... rellenar pnls ...
|
|
|
|
double mean = fn::ds::stats_mean(pnls.data(), N);
|
|
double std = fn::ds::stats_std (pnls.data(), N);
|
|
|
|
// CI 95% via percentiles 2.5 / 97.5
|
|
std::vector<double> sorted(N);
|
|
fn::ds::stats_sort(pnls.data(), N, sorted.data());
|
|
double p025 = fn::ds::stats_quantile_sorted(sorted.data(), N, 0.025);
|
|
double p975 = fn::ds::stats_quantile_sorted(sorted.data(), N, 0.975);
|
|
```
|
|
|
|
## Notas
|
|
|
|
- El convenio R type-7 para quantiles es el mismo que numpy default (`linear`) y matplotlib. Pasar tests numericos contra numpy debe matchear bit-exacto.
|
|
- `sample=true` (default) coincide con `np.var(x, ddof=1)` y `pd.DataFrame.var()`.
|
|
- Para datasets enormes que no caben en RAM, usar `gpu_reduce` (GPU) — esta libreria es CPU-side.
|