scipy.stats._result_classes.EmpiricalDistributionFunction.confidence_interval#

EmpiricalDistributionFunction.confidence_interval(confidence_level=0.95, *, method='linear')[source]#

Compute a confidence interval around the CDF/SF point estimate

Parameters:

confidence_levelfloat, default: 0.95: Confidence level for the computed confidence interval
methodstr, {“linear”, “log-log”}: Method used to compute the confidence interval. Options are “linear” for the conventional Greenwood confidence interval (default) and “log-log” for the “exponential Greenwood”, log-negative-log-transformed confidence interval.

Returns:

ciConfidenceInterval: An object with attributes low and high, instances of EmpiricalDistributionFunction that represent the lower and upper bounds (respectively) of the confidence interval.

Notes

Confidence intervals are computed according to the Greenwood formula (method='linear') or the more recent “exponential Greenwood” formula (method='log-log') as described in [1]. The conventional Greenwood formula can result in lower confidence limits less than 0 and upper confidence limits greater than 1; these are clipped to the unit interval. NaNs may be produced by either method; these are features of the formulas.

References

[1]

Sawyer, Stanley. “The Greenwood and Exponential Greenwood Confidence Intervals in Survival Analysis.” https://www.math.wustl.edu/~sawyer/handouts/greenwood.pdf