output_analysis.confidence_interval_method()

Determine the minimum number of simulation replications required to achieve

Usage

Source

output_analysis.confidence_interval_method(
    replications,
    alpha=0.05,
    desired_precision=0.1,
    min_rep=5,
    decimal_places=2
)

a target precision in the confidence interval of one or several performance metrics.

This function applies the confidence interval method: it identifies the smallest replication count where the relative half-width of the confidence interval is less than the specified desired_precision for each metric.

Parameters

replications: Union[
    pd.Series,
    pd.DataFrame,
    Sequence[float],
    Sequence[Sequence[float]],
    Dict[str, Sequence[float]],
]


Replication results for one or more performance metrics. Accepted formats:

• pd.Series or 1D list/numpy array → single metric
• pd.DataFrame → multiple metrics in columns
• dict[str, list/array/Series] → {metric_name: replications}
• list of lists / numpy arrays / Series → multiple metrics unnamed Each inner sequence/Series/numpy array must contain numeric replication results in the order they were generated.

alpha: Optional[float] = 0.05

Significance level for confidence interval calculations (CI level = 100 * (1 - alpha) %).

desired_precision: Optional[float] = 0.1

Target CI half-width precision (i.e. percentage deviation of the confidence interval from the mean).

min_rep: Optional[int] = 5

Minimum number of replications to consider before evaluating precision. Helps avoid unstable early results.

decimal_places: Optional[int] = 2

Number of decimal places to round values in the returned results table.

Returns

- Single-metric input → tuple (n_reps, results_df)

- Multi-metric input → dict:

{metric_name: (n_reps, results_df)}

Where

n_reps : int The smallest number of replications achieving the desired precision. Returns -1 if precision is never reached. results_df : pandas.DataFrame Summary statistics at each replication: “Mean”, “Cumulative Mean”, “Standard Deviation”, “Lower Interval”, “Upper Interval”, “% deviation”

Warns

UserWarning

Issued per metric if the desired precision is never reached.