假设检验
除非另有说明,每个检验返回 LIST(DOUBLE) 形如 [统计量, p 值]。样本参数是 LIST(DOUBLE)。
代码类参数(alternative、nan_policy、method)用 DOUBLE 字面量,与 statrs 的枚举对应。
sr_ttest_onesample(sample, popmean, alternative, nan_policy)
签名:sr_ttest_onesample(sample DOUBLE[], popmean DOUBLE, alternative DOUBLE, nan_policy DOUBLE) -> LIST(DOUBLE)
单样本 t 检验,与 popmean 比较。返回 [t 统计量, p 值]。
sample:观测值列表popmean:待检验的总体均值alternative:1.0 = 双侧,2.0 = less,3.0 = greaternan_policy:1.0 = propagate,2.0 = omit,3.0 = raise(对应 statrs 的枚举变体)
SELECT sr_ttest_onesample([1.0, 2.0, 3.0, 4.0, 5.0], 3.0, 1.0, 1.0)
-- [0.0, 1.0]
sr_chisquare(observed, expected, ddof)
签名:sr_chisquare(observed UBIGINT[], expected DOUBLE[], ddof UBIGINT) -> LIST(DOUBLE)
卡方拟合优度检验。expected 传 NULL 表示均匀期望;ddof 传 NULL 默认为 0。
返回 [卡方统计量, p 值]。
SELECT sr_chisquare([16::UBIGINT, 18::UBIGINT, 16::UBIGINT, 14::UBIGINT, 12::UBIGINT, 12::UBIGINT], NULL, NULL)
-- [2.0, 0.8491450360846101]
sr_f_oneway(groups, nan_policy)
签名:sr_f_oneway(groups DOUBLE[][], nan_policy DOUBLE) -> LIST(DOUBLE)
单因素 ANOVA,多组样本。返回 [F 统计量, p 值]。
SELECT sr_f_oneway([[1.0, 3.0, 5.0], [2.0, 4.0, 8.0]], 1.0)
-- [0.6249999999999984, 0.47342736525713647]
sr_fishers_exact(table, alternative)
签名:sr_fishers_exact(table UBIGINT[], alternative DOUBLE) -> DOUBLE
2x2 列联表(4 个 UBIGINT,行主序)的 Fisher 精确检验 p 值。返回单个 DOUBLE,不是 LIST。
alternative:1.0 = 双侧,2.0 = less,3.0 = greater
SELECT sr_fishers_exact([1::UBIGINT, 2::UBIGINT, 3::UBIGINT, 4::UBIGINT], 1.0)
-- 1.0
sr_fishers_exact_with_odds_ratio(table, alternative)
签名:sr_fishers_exact_with_odds_ratio(table UBIGINT[], alternative DOUBLE) -> LIST(DOUBLE)
与 sr_fishers_exact 同,返回 [odds_ratio, p 值]。
SELECT sr_fishers_exact_with_odds_ratio([1::UBIGINT, 2::UBIGINT, 3::UBIGINT, 4::UBIGINT], 2.0)
-- [0.6666666666666666, 0.6666666666666666]
sr_ks_twosample(sample1, sample2, alternative, method)
签名:sr_ks_twosample(sample1 DOUBLE[], sample2 DOUBLE[], alternative DOUBLE, method DOUBLE) -> LIST(DOUBLE)
两样本 Kolmogorov-Smirnov 检验。返回 [D 统计量, p 值]。
alternative:1.0 = less,2.0 = greater,3.0 = 双侧精确,4.0 = 双侧渐近method:模式选择,一般用 1.0(自动)
SELECT sr_ks_twosample([0.1, 0.2, 0.3], [0.4, 0.5, 0.6], 4.0, 1.0)
-- [1.0, 0.09956184831478034]
sr_ks_onesample(x, dist, params, method, nan)
签名:sr_ks_onesample(x DOUBLE[], dist VARCHAR, params DOUBLE[], method DOUBLE, nan DOUBLE) -> DOUBLE[]
单样本 Kolmogorov-Smirnov 检验:把 LIST(DOUBLE) 样本与命名分布(normal / lognormal / exponential / gumbel / weibull / uniform)及其参数 LIST 比较,返回 LIST [KS 统计量, p 值];method 取 1 less、2 greater、3 双侧精确、4 双侧渐近、5 双侧近似。
SELECT sr_ks_onesample([1.0, 2.0, 3.0, 4.0], 'normal', [2.5, 1.0], 4.0, 1.0)
-- [0.19146246127401312, 0.9985479186818191]
sr_mannwhitneyu(sample1, sample2, alternative, method)
签名:sr_mannwhitneyu(sample1 DOUBLE[], sample2 DOUBLE[], alternative DOUBLE, method DOUBLE) -> LIST(DOUBLE)
Mann-Whitney U 检验。返回 [U 统计量, p 值]。
alternative:1.0 = 双侧,2.0 = less,3.0 = greatermethod:1.0 = 自动,2.0 = 精确,3.0 = 渐近含连续性校正,4.0 = 渐近不含
SELECT sr_mannwhitneyu([1.0, 2.0, 3.0], [4.0, 5.0, 6.0], 2.0, 2.0)
-- [0.0, 0.95]
sr_skewtest(sample, alternative, nan_policy)
签名:sr_skewtest(sample DOUBLE[], alternative DOUBLE, nan_policy DOUBLE) -> LIST(DOUBLE)
偏度 z 检验。返回 [z 统计量, p 值]。
alternative:1.0 = 双侧,2.0 = less(负偏),3.0 = greater(正偏)nan_policy:与 t 检验同
SELECT sr_skewtest([1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0], 1.0, 1.0)
-- [1.018464355396213, 0.30845733197774927]
sr_anderson_darling(sample, distribution, params)
签名:sr_anderson_darling(sample DOUBLE[], distribution VARCHAR, params DOUBLE[]) -> LIST(DOUBLE)
对指定分布的 Anderson-Darling 拟合优度检验。返回 [A², 5% 临界值]。
distribution:'normal'、'lognormal'、'exponential'、'gumbel'、'weibull'、'uniform'之一params:分布参数列表(正态是[mean, std_dev])
SELECT sr_anderson_darling([1.0, 2.0, 3.0, 4.0, 5.0], 'normal', [3.0, 1.5])
-- [0.15491765936161173, 0.8959843708415298]
从列构造样本 LIST
数据在表里时用 list(x) 喂给检验:
SELECT sr_ttest_onesample(list(x), 3.0, 1.0, 1.0)
FROM (VALUES (1.0), (2.0), (3.0), (4.0), (5.0)) t(x)
-- [0.0, 1.0]
非法 code 报查询错误
SELECT sr_ttest_onesample([1.0, 2.0], 3.0, 9.0, 1.0)
-- error: sr_ttest_onesample: the alternative must be 1 = two-sided, 2 = less, 3 = greater (statrs' Alternative), got 9