"""Original independent reconstruction of primary rank/HC3 models from supplied feature outputs.
No submitted module is imported. This does not redetect physiological peaks from raw signals.
Usage: python independent_regression.py <bundle> <external-cache>
"""
import csv,json,math,sys
from pathlib import Path
import numpy as np
from scipy import stats
b,c=map(Path,sys.argv[1:3]);correct_gender='--correct-gender' in sys.argv[3:];pointers=json.loads((b/'data/external.json').read_text());paths={x['path']:c/x['sha256'][7:] for x in pointers}
def read(p):
 with p.open(newline='') as f:return list(csv.DictReader(f,delimiter='\t' if p.name not in ['features.csv'] else ','))
def tab(path):
 with paths[path].open(newline='') as f:return {x['participant_id']:x for x in csv.DictReader(f,delimiter='\t')}
q=tab('data/ds003823/phenotype/Emotion_questionnaires_summary.tsv');dem=tab('data/ds003823/phenotype/Demographics.tsv')
with (b/'results/features.csv').open(newline='') as f:features=list(csv.DictReader(f))
def number(x):
 try:return float(x)
 except (ValueError,TypeError):return math.nan
def mean2(x,y):
 a=[z for z in [number(x),number(y)] if math.isfinite(z)];return sum(a)/len(a) if a else math.nan
rows=[]
for x in features:
 pid=x['participant_id']
 if pid not in q:continue
 row={k:number(v) for k,v in x.items() if k not in ['participant_id','error','hr_ok','sbp_ok']};z=q[pid];row.update({'cesd':mean2(z.get('cesd_Total_w1'),z.get('cesd_Total_w2')),'older':float(z.get('age_group')=='OA'),'female':float(dem.get(pid,{}).get('gender')==('female' if correct_gender else 'F')),'hr_ok':x['hr_ok']=='True','sbp_ok':x['sbp_ok']=='True'})
 if not math.isfinite(row['cesd']):continue
 for sig in ['hr','sbp']:
  row[sig+'_ar1_change']=row[sig+'_ar1_recovery']-row[sig+'_ar1_baseline'];row[sig+'_reactivity']=row[sig+'_stress']-row[sig+'_baseline']
 rows.append(row)
def reg(data,out,covs,positive=True):
 use=[z for z in data if all(math.isfinite(z[k]) for k in [out,'cesd']+covs)];raw=np.array([[z[k] for k in ['cesd']+covs] for z in use]);cols=[]
 for v in raw.T:cols.append(stats.rankdata(v) if len(set(v))>2 else v)
 X=np.column_stack([np.ones(len(use))]+cols);y=stats.rankdata([z[out] for z in use]);coef=np.linalg.lstsq(X,y,rcond=None)[0];pinv=np.linalg.pinv(X,rcond=1e-15);bread=pinv@pinv.T;res=y-X@coef;hat=np.sum((X@bread)*X,axis=1);meat=X.T@(((res/(1-hat))**2)[:,None]*X);V=bread@meat@bread;t=coef[1]/math.sqrt(V[1,1]);scale=np.std(X[:,1])/np.std(y);beta=coef[1]*scale;se=math.sqrt(V[1,1])*scale;return {'n':len(use),'beta_std':round(float(beta),4),'z_unrounded':float(t),'p_one_sided':float(stats.norm.sf(t) if positive else stats.norm.cdf(t)),'marginal_95pct_interval_std':[float(beta-1.959963984540054*se),float(beta+1.959963984540054*se)]}
P={}
for sig in ['hr','sbp']:
 data=[z for z in rows if z[sig+'_ok']]
 P[sig+'_late_recovery']=reg(data,sig+'_late_recovery',[sig+'_baseline',sig+'_stress','older','female']);P[sig+'_ar1_change']=reg(data,sig+'_ar1_change',['older','female'])
run=0
for i,k in enumerate(sorted(P,key=lambda k:P[k]['p_one_sided'])):
 run=max(run,min(1,(len(P)-i)*P[k]['p_one_sided']));P[k]['p_holm']=round(run,5)
d=[z for z in rows if z['sbp_ok']];hi=[z['sbp_late_recovery'] for z in d if z['cesd']>=16];lo=[z['sbp_late_recovery'] for z in d if z['cesd']<16];u=stats.mannwhitneyu(hi,lo)
S={'sbp_late_recovery_cesd16_auc':{'auc':round(float(u.statistic/(len(hi)*len(lo))),4),'p_two_sided':float(u.pvalue),'n_high':len(hi),'n_low':len(lo)},'hr_reactivity_blunting':reg([z for z in rows if z['hr_ok']],'hr_reactivity',['hr_baseline','older','female'],False)}
out={'scope':'Independent rank-regression/HC3 arithmetic from supplied feature outputs; not independent peak extraction','gender_code':('female' if correct_gender else 'F'),'female_covariate_ones':sum(z['female'] for z in rows),'n_with_cesd':len(rows),'n_hr_ok':sum(z['hr_ok'] for z in rows),'n_sbp_ok':sum(z['sbp_ok'] for z in rows),'primary':P,'secondary':S};print(json.dumps(out,indent=2))
