ppv and ppv_upper, ppv_lower mismatch

[onishchenkod]

When plotting ppv, ppv_lower and ppv_upper as generated by zedstat, I get the following mismatch (as shown in attached plot).

ax.plot(PRC['tpr'],
            PRC['ppv'],
            label = "%s: (%s)" % (LABEL, PREVAL_LABEL) if LABEL else PREVAL_LABEL,
            color = COLOR)
plt.fill_between(
    x = PRC['tpr'],
    y1 = PRC['ppv_upper'],
    y2 = PRC['ppv_lower'],
    color = COLOR,
    alpha = CB_ALPHA
)

The code used to generate the plotted curve is

    fpr, tpr, thresholds = roc_curve(DF[target_col], DF[risk_col])
    ROC = pd.DataFrame({
        'fpr': fpr,
        'tpr': tpr,
        'threshold': thresholds,
    })
    zt=zedstat.processRoc(df=ROC,
                          order=3,
                          total_samples=DF.shape[0],
                          positive_samples=DF[target_col].value_counts()[1],
                          alpha=0.05,
                          prevalence=DF[target_col].value_counts(normalize = True)[1] if not preval else preval)
    zt.smooth(STEP=0.001)
    zt.allmeasures(interpolate=True)
    zt.usample(precision=3)
    zt.getBounds()
    if bounds:
        df_=zt.get()
        df_u=zt.df_lim['U']
        df_l=zt.df_lim['L']
        df_=df_.join(df_u,rsuffix='_upper').join(df_l,rsuffix='_lower')
        CURVE = df_.reset_index()
        zt.df = CURVE
    return zt

this is how the same curve's plot looks like when I recompute the ppv by

PRC['ppv'] = (PRC['ppv_upper'] + PRC['ppv_lower'])/2

Either ppv computation or computation of bounds is not done correctly here.

[onishchenkod]

zedstat version used: zedstat==0.0.128

[onishchenkod]

PPV bounds match those of TPR bounds