Skip to content

openavmkit.vertical_equity_study

Vertical equity analysis.

Implements :class:VerticalEquityStudy, which measures whether high-value parcels and low-value parcels are valued with the same accuracy — the "vertical equity" question. Computes PRD and PRB (with bootstrap confidence intervals) plus per-quantile median ratios.

Together with :mod:openavmkit.horizontal_equity_study and :mod:openavmkit.ratio_study, this module forms OpenAVMKit's IAAO-aligned equity analysis suite.

VerticalEquityStudy 

VerticalEquityStudy(df_sales_in, field_sales, field_prediction, field_location, confidence_interval=0.95, iterations=10000, seed=777)

Perform vertical equity analysis and summarize the results.

Attributes:

Name Type Description
rows int

Total number of rows in the input DataFrame.
confidence_interval float

The confidence interval (e.g. 0.95 for 95% confidence)
prd ConfidenceStat

The price-related differential, with confidence intervals
prb ConfidenceStat

The price-related bias, with confidence intervals
quantiles DataFrame

A dataframe containing the median ratio, with confidence intervals, of all ten price quantile tiers
Source code in openavmkit/vertical_equity_study.py 
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
def __init__(
    self,
    df_sales_in: pd.DataFrame,
    field_sales: str,
    field_prediction: str,
    field_location: str,
    confidence_interval : float = 0.95,
    iterations: int = 10000,
    seed : int = 777
):
    df_sales = df_sales_in.copy()

    n = len(df_sales)
    self.rows = n
    self.confidence_interval = confidence_interval

    # Calculate PRD and PRB
    #----------------------

    predictions = df_sales[field_prediction].to_numpy()
    sales = df_sales[field_sales].to_numpy()

    results = calc_ratio_stats_bootstrap(predictions, sales, confidence_interval, iterations=iterations, seed=seed)

    prb_point, prb_low, prb_high = calc_prb(predictions, sales, confidence_interval)

    self.prb = ConfidenceStat(prb_point, confidence_interval, prb_low, prb_high)

    # Calculate quantiles (directly from price)
    #------------------------------------------

    df_sales["quantile"] = _calc_quantiles(df_sales_in, field_sales)
    df = _assemble_quantile_df(df_sales, field_sales, field_prediction, confidence_interval, iterations, seed)
    self.quantiles = df
    # Calculate quantiles (grouped price)
    #------------------------------------------

    if field_location is not None and field_location in df_sales_in.columns:
        df_sales["quantile"] = _calc_grouped_quantiles(df_sales_in, field_sales, field_location)
        df = _assemble_quantile_df(df_sales, field_sales,  field_prediction, confidence_interval, iterations, seed)
        self.grouped_quantiles = df
    else:
        warnings.warn(f"VerticalEquityStudy: could not find location field, \"{field_location}\" in sales dataframe")
        self.grouped_quantiles = None