Categorical Statistics Functions

Return to Homepage

Categorical Statistics Functions

gval.statistics.categorical_stat_funcs.accuracy(tp: Number, tn: Number, fp: Number, fn: Number) → float

Computes accuracy

Parameters:

• tp (Number) – Count reflecting true positive

• tn (Number) – Count reflecting true negative

• fp (Number) – Count reflecting false positive

• fn (Number) – Count reflecting false negative

Returns:

Accuracy from 0 to 1

Return type:

float

References

[1]

Accuracy and Precision

gval.statistics.categorical_stat_funcs.balanced_accuracy(tp: Number, tn: Number, fp: Number, fn: Number) → float

Computes Balanced Accuracy

Parameters:

• tp (Number) – Count reflecting true positive

• tn (Number) – Count reflecting true negative

• fp (Number) – Count reflecting false positive

• fn (Number) – Count reflecting false negative

Returns:

Balanced Accuracy from 0 to 1

Return type:

float

References

[1]

Balanced Accuracy

gval.statistics.categorical_stat_funcs.critical_success_index(tp: Number, fp: Number, fn: Number) → float

Computes critical success index

https://www.weather.gov/media/erh/ta2004-03.pdf

Parameters:

• tp (Number) – Count reflecting true positive

• fp (Number) – Count reflecting false positive

• fn (Number) – Count reflecting false negative

Returns:

Critical success index from 0 to 1

Return type:

float

References

[1]

Critical Success Index

gval.statistics.categorical_stat_funcs.equitable_threat_score(tp: Number, tn: Number, fp: Number, fn: Number) → float

Computes Equitable Threat Score (Gilbert Score)

Parameters:

• tp (Number) – Count reflecting true positive

• tn (Number) – Count reflecting true negative

• fp (Number) – Count reflecting false positive

• fn (Number) – Count reflecting false negative

Returns:

Equitable threat score from -1/3 to 1

Return type:

float

References

[1]

Equitable Threat Score

gval.statistics.categorical_stat_funcs.f_score(tp: Number, fp: Number, fn: Number) → float

Computes F-score AKA harmonic mean of precision and sensitivity

Parameters:

• tp (Number) – Count reflecting true positive

• fp (Number) – Count reflecting false positive

• fn (Number) – Count reflecting false negative

Returns:

F-score from 0 to 1

Return type:

float

References

[1]

F-score

gval.statistics.categorical_stat_funcs.false_discovery_rate(tp: Number, fp: Number) → float

Computes false discovery rate

Parameters:

• tp (Number) – Count reflecting true positive

• fp (Number) – Count reflecting false positive

Returns:

False discovery rate from 0 to 1

Return type:

float

References

[1]

False Discovery Rate

gval.statistics.categorical_stat_funcs.false_negative_rate(tp: Number, fn: Number) → float

Computes false negative rate

Parameters:

• tp (Number) – Count reflecting true positive

• fn (Number) – Count reflecting false negative

Returns:

False negative rate from 0 to 1

Return type:

float

References

[1]

Type I and Type II Error

gval.statistics.categorical_stat_funcs.false_omission_rate(tn: Number, fn: Number) → float

Computes false omission rate

Parameters:

• tn (Number) – Count reflecting true negative

• fn (Number) – Count reflecting false negative

Returns:

False omission rate from 0 to 1

Return type:

float

References

[1]

Positive and Negative Predictive Values

gval.statistics.categorical_stat_funcs.false_positive_rate(tn: Number, fp: Number) → float

Computes false positive rate AKA fall-out

Parameters:

• tn (Number) – Count reflecting true negative

• fp (Number) – Count reflecting false positive

Returns:

False positive rate from 0 to 1

Return type:

float

References

[1]

Type I and Type II Errors

gval.statistics.categorical_stat_funcs.fowlkes_mallows_index(tp: Number, fp: Number, fn: Number) → float

Computes Fowlkes-Mallows index

Parameters:

• tp (Number) – Count reflecting true positive

• fp (Number) – Count reflecting false positive

• fn (Number) – Count reflecting false negative

Returns:

Correlation coefficient from -1 to 1

Return type:

float

References

[1]

Fowlkes-Mallows Index

gval.statistics.categorical_stat_funcs.matthews_correlation_coefficient(tp: Number, tn: Number, fp: Number, fn: Number) → float

Computes matthews correlation coefficient, accounting for accuracy and precision of both true positives and true negatives AKA Phi Coefficient

Parameters:

• tp (Number) – Count reflecting true positive

• tn (Number) – Count reflecting true negative

• fp (Number) – Count reflecting false positive

• fn (Number) – Count reflecting false negative

Returns:

Correlation coefficient from -1 to 1

Return type:

float

References

[1]

Matthews Correlation Coefficient

gval.statistics.categorical_stat_funcs.negative_likelihood_ratio(tp: Number, tn: Number, fp: Number, fn: Number) → float

Computes negative likelihood ratio

Parameters:

• tp (Number) – Count reflecting true positive

• tn (Number) – Count reflecting true negative

• fp (Number) – Count reflecting false positive

• fn (Number) – Count reflecting false negative

Returns:

Negative likelihood from 1 to infinity

Return type:

float

References

[1]

Likelihood Ratios

gval.statistics.categorical_stat_funcs.negative_predictive_value(tn: Number, fn: Number) → float

Computes negative predictive value

Parameters:

• tn (Number) – Count reflecting true negative

• fn (Number) – Count reflecting false negative

Returns:

Negative predictive value from 0 to 1

Return type:

float

References

[1]

Sensitivity and Specificity

gval.statistics.categorical_stat_funcs.overall_bias(tp: Number, fp: Number, fn: Number) → float

Computes the degree of correspondence between the mean forecast and the mean observation.

Parameters:

• tp (Number) – Count reflecting true positive

• fp (Number) – Count reflecting false positive

• fn (Number) – Count reflecting false negative

Returns:

Overall Bias

Return type:

float

References

[1]

Forecast Verification Glossary

gval.statistics.categorical_stat_funcs.positive_likelihood_ratio(tp: Number, tn: Number, fp: Number, fn: Number) → float

Computes positive likelihood ratio

Parameters:

• tp (Number) – Count reflecting true positive

• tn (Number) – Count reflecting true negative

• fp (Number) – Count reflecting false positive

• fn (Number) – Count reflecting false negative

Returns:

Positive likelihood rate from 1 to infinity

Return type:

float

References

[1]

Likelihood Ratios

gval.statistics.categorical_stat_funcs.positive_predictive_value(tp: Number, fp: Number) → float

Computes positive predictive value AKA precision

Parameters:

• tp (Number) – Count reflecting true positive

• fp (Number) – Count reflecting false positive

Returns:

Positive predictive value from 0 to 1

Return type:

float

References

[1]

Sensitivity and Specificity

gval.statistics.categorical_stat_funcs.prevalence(tp: Number, tn: Number, fp: Number, fn: Number) → float

Computes prevalence

Parameters:

• tp (Number) – Count reflecting true positive

• tn (Number) – Count reflecting true negative

• fp (Number) – Count reflecting false positive

• fn (Number) – Count reflecting false negative

Returns:

Prevalence from 0 to 1

Return type:

float

References

[1]

Prevalence

gval.statistics.categorical_stat_funcs.prevalence_threshold(tp: Number, tn: Number, fp: Number, fn: Number) → float

Computes prevalence threshold

Parameters:

• tp (Number) – Count reflecting true positive

• tn (Number) – Count reflecting true negative

• fp (Number) – Count reflecting false positive

• fn (Number) – Count reflecting false negative

Returns:

Prevalence threshold from 0 to 1

Return type:

float

References

[1]

Prevalence Threshold

gval.statistics.categorical_stat_funcs.true_negative_rate(tn: Number, fp: Number) → float

Computes true negative rate, AKA specificity, selectivity

Parameters:

• tn (Number) – Count reflecting true negative

• fp (Number) – Count reflecting false positive

Returns:

True negative rate from 0 to 1

Return type:

float

References

[1]

Sensitivity and Specificity

gval.statistics.categorical_stat_funcs.true_positive_rate(tp: Number, fn: Number) → float

Computes true positive rate, AKA sensitivity, recall, hit rate

Parameters:

• tp (Number) – Count reflecting true positive

• fn (Number) – Count reflecting false negative

Returns:

True positive rate from 0 to 1

Return type:

float

References

[1]

Sensitivity and Specificity