From a9be767761f7ff2e257c3748e6afd93b296a0b93 Mon Sep 17 00:00:00 2001
From: TimLai666 <43640816+TimLai666@users.noreply.github.com>
Date: Sat, 14 Sep 2024 14:07:19 +0800
Subject: [PATCH] =?UTF-8?q?=E9=80=90=E6=AD=A5=E6=B7=98=E6=B1=B0=E9=AB=98?=
 =?UTF-8?q?=E7=B2=BE=E5=BA=A6=E6=A8=A1=E5=BC=8F?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 datalist.go                | 72 +++++++++++++-------------------------
 insyra_test.go             |  2 +-
 stats/anova.go             |  6 ++--
 stats/correlation.go       |  4 +--
 stats/linear_regression.go |  4 +--
 stats/moments.go           |  2 +-
 stats/skewness.go          |  2 +-
 stats/ttest.go             |  8 ++---
 8 files changed, 39 insertions(+), 61 deletions(-)

diff --git a/datalist.go b/datalist.go
index c910074..142a100 100644
--- a/datalist.go
+++ b/datalist.go
@@ -71,10 +71,11 @@ type IDataList interface {
 	Upper() *DataList
 	Lower() *DataList
 	Capitalize() *DataList
+	// Statistics
 	Sum() interface{}
 	Max() interface{}
 	Min() interface{}
-	Mean(highPrecision ...bool) interface{}
+	Mean() float64
 	WeightedMean(weights interface{}) interface{}
 	GMean() interface{}
 	Median(highPrecision ...bool) interface{}
@@ -407,7 +408,7 @@ func (dl *DataList) ReplaceAll(oldValue, newValue interface{}) {
 
 // ReplaceOutliers replaces outliers in the DataList with the specified replacement value (e.g., mean, median).
 func (dl *DataList) ReplaceOutliers(stdDevs float64, replacement float64) *DataList {
-	mean := dl.Mean(false).(float64)
+	mean := dl.Mean()
 	stddev := dl.Stdev(false).(float64)
 	threshold := stdDevs * stddev
 
@@ -693,7 +694,7 @@ func (dl *DataList) ClearOutliers(stdDevs float64) *DataList {
 		go dl.updateTimestamp()
 	}()
 
-	mean := dl.Mean(false).(float64)
+	mean := dl.Mean()
 	stddev := dl.Stdev(false).(float64)
 	threshold := stdDevs * stddev
 
@@ -750,7 +751,7 @@ func (dl *DataList) Standardize() *DataList {
 		go reorganizeMemory(dl)
 		go dl.updateTimestamp()
 	}()
-	mean := dl.Mean(false).(float64)
+	mean := dl.Mean()
 	stddev := dl.Stdev(false).(float64)
 	dl.mu.Lock()
 	for i, v := range dl.Data() {
@@ -774,7 +775,7 @@ func (dl *DataList) FillNaNWithMean() *DataList {
 	}()
 	dlclone := dl.Clone()
 	dlNoNaN := dlclone.ClearNaNs()
-	mean := dlNoNaN.Mean(false).(float64)
+	mean := dlNoNaN.Mean()
 	dl.mu.Lock()
 	for i, v := range dl.Data() {
 		vfloat := conv.ParseF64(v)
@@ -1137,56 +1138,33 @@ func (dl *DataList) Min() interface{} {
 }
 
 // Mean calculates the arithmetic mean of the DataList.
-// If highPrecision is true, it will calculate using big.Rat for high precision.
-// Otherwise, it calculates using float64.
-// Returns nil if the DataList is empty or if an invalid number of parameters is provided.
-func (dl *DataList) Mean(highPrecision ...bool) interface{} {
+// Returns math.NaN() if the DataList is empty or if no elements can be converted to float64.
+func (dl *DataList) Mean() float64 {
+	mean := math.NaN()
 	if len(dl.data) == 0 {
-		LogWarning("DataList.Mean(): DataList is empty, returning nil.")
-		return nil
-	}
-
-	// 檢查參數數量
-	if len(highPrecision) > 1 {
-		LogWarning("DataList.Mean(): Too many arguments, returning nil.")
-		return nil
-	}
-
-	// 默認使用普通模式（float64），若有參數則使用參數設定
-	highPrecisionMode := false
-	if len(highPrecision) == 1 {
-		highPrecisionMode = highPrecision[0]
-	}
-
-	if highPrecisionMode {
-		sum := new(big.Rat)
-		count := big.NewRat(int64(len(dl.data)), 1)
-
-		for _, v := range dl.data {
-			if val, ok := ToFloat64Safe(v); ok {
-				ratValue := new(big.Rat).SetFloat64(val)
-				sum.Add(sum, ratValue)
-			} else {
-				LogWarning("DataList.Mean(): Data types cannot be compared, returning nil.")
-				return nil
-			}
-		}
-
-		mean := new(big.Rat).Quo(sum, count)
+		LogWarning("DataList.Mean(): DataList is empty.")
 		return mean
 	}
 
-	// 普通模式（float64）
 	var sum float64
+	var count int
 	for _, v := range dl.data {
 		if val, ok := ToFloat64Safe(v); ok {
 			sum += val
+			count++
 		} else {
-			LogWarning("DataList.Mean(): Data types cannot be compared, returning nil.")
-			return nil
+			LogWarning("DataList.Mean(): Element %v cannot be converted to float64, skipping.", val)
+			// 跳过无法转换的元素
+			continue
 		}
 	}
-	mean := sum / float64(len(dl.data))
+
+	if count == 0 {
+		LogWarning("DataList.Mean(): No elements could be converted to float64.")
+		return mean
+	}
+
+	mean = sum / float64(count)
 	return mean
 }
 
@@ -1434,7 +1412,7 @@ func (dl *DataList) Var(highPrecision ...bool) interface{} {
 
 	if useHighPrecision {
 		// 使用 big.Rat 進行高精度計算
-		mean := dl.Mean(true).(*big.Rat)
+		mean := new(big.Rat).SetFloat64(dl.Mean())
 		denominator := new(big.Rat).SetFloat64(n - 1)
 		if denominator.Cmp(big.NewRat(0, 1)) == 0 {
 			LogWarning("DataList.Var(): Denominator is 0, returning nil.")
@@ -1457,7 +1435,7 @@ func (dl *DataList) Var(highPrecision ...bool) interface{} {
 	}
 
 	// 普通模式使用 float64 計算
-	mean := dl.Mean(false).(float64)
+	mean := dl.Mean()
 	denominator := n - 1
 	if denominator == 0 {
 		LogWarning("DataList.Var(): Denominator is 0, returning nil.")
@@ -1494,7 +1472,7 @@ func (dl *DataList) VarP(highPrecision ...bool) interface{} {
 
 	if useHighPrecision {
 		// 使用高精度计算
-		mean := dl.Mean(true).(*big.Rat)
+		mean := new(big.Rat).SetFloat64(dl.Mean())
 		numerator := new(big.Rat)
 		for i := 0; i < len(dl.data); i++ {
 			xi := new(big.Rat).SetFloat64(ToFloat64(dl.data[i]))
diff --git a/insyra_test.go b/insyra_test.go
index 3009189..3aafc23 100644
--- a/insyra_test.go
+++ b/insyra_test.go
@@ -162,7 +162,7 @@ func TestMean(t *testing.T) {
 	dl := NewDataList(1, 2, 3, 4)
 	mean := dl.Mean()
 
-	if v, ok := mean.(float64); !ok || !float64Equal(v, 2.5) {
+	if !float64Equal(mean, 2.5) {
 		t.Errorf("Expected mean 2.5, got %v", mean)
 	}
 }
diff --git a/stats/anova.go b/stats/anova.go
index f705997..f47f76d 100644
--- a/stats/anova.go
+++ b/stats/anova.go
@@ -45,14 +45,14 @@ func OneWayANOVA_WideFormat(dataTable insyra.IDataTable) *OneWayANOVAResult {
 		func() {
 			SSB = 0.0
 			for _, group := range groups {
-				groupMean := group.Mean().(float64)
+				groupMean := group.Mean()
 				SSB += float64(group.Len()) * math.Pow(groupMean-totalMean, 2)
 			}
 		},
 		func() {
 			SSW = 0.0
 			for _, group := range groups {
-				groupMean := group.Mean().(float64)
+				groupMean := group.Mean()
 				for i := 0; i < group.Len(); i++ {
 					value, _ := group.Get(i).(float64)
 					SSW += math.Pow(value-groupMean, 2)
@@ -305,7 +305,7 @@ func RepeatedMeasuresANOVA_WideFormat(dataTable insyra.IDataTable) *RepeatedMeas
 	ssBetweenFunc := func() {
 		ssBetween = 0.0
 		for i := 0; i < rowNum; i++ {
-			conditionMean := dataTable.GetRow(i).Mean().(float64)
+			conditionMean := dataTable.GetRow(i).Mean()
 			ssBetween += float64(colNum) * math.Pow(conditionMean-grandMean, 2)
 		}
 	}
diff --git a/stats/correlation.go b/stats/correlation.go
index 6101430..8b57c31 100644
--- a/stats/correlation.go
+++ b/stats/correlation.go
@@ -25,8 +25,8 @@ const (
 // Covariance calculates the covariance between two datasets.
 // Always returns *big.Rat.
 func Covariance(dlX, dlY insyra.IDataList) *big.Rat {
-	meanX := dlX.Mean(true).(*big.Rat)
-	meanY := dlY.Mean(true).(*big.Rat)
+	meanX := new(big.Rat).SetFloat64(dlX.Mean())
+	meanY := new(big.Rat).SetFloat64(dlY.Mean())
 
 	cov := new(big.Rat)
 	for i := 0; i < dlX.Len(); i++ {
diff --git a/stats/linear_regression.go b/stats/linear_regression.go
index 8ae3e86..5fb4c1e 100644
--- a/stats/linear_regression.go
+++ b/stats/linear_regression.go
@@ -30,8 +30,8 @@ func LinearRegression(dlX, dlY insyra.IDataList) *LinearRegressionResult {
 	}
 
 	// 計算 X 和 Y 的均值
-	meanX := dlX.Mean(true).(*big.Rat)
-	meanY := dlY.Mean(true).(*big.Rat)
+	meanX := new(big.Rat).SetFloat64(dlX.Mean())
+	meanY := new(big.Rat).SetFloat64(dlY.Mean())
 
 	// 初始化變量
 	numerator := new(big.Rat)
diff --git a/stats/moments.go b/stats/moments.go
index 162d827..c89605e 100644
--- a/stats/moments.go
+++ b/stats/moments.go
@@ -20,7 +20,7 @@ func CalculateMoment(dl insyra.IDataList, n int, central bool) *big.Rat {
 	// 初始化均值
 	mean := new(big.Rat)
 	if central {
-		mean = dl.Mean(true).(*big.Rat) // 計算均值
+		mean = new(big.Rat).SetFloat64(dl.Mean()) // 計算均值
 	}
 
 	// 初始化 n 階矩
diff --git a/stats/skewness.go b/stats/skewness.go
index 09b92ca..060fce2 100644
--- a/stats/skewness.go
+++ b/stats/skewness.go
@@ -62,7 +62,7 @@ func Skewness(sample interface{}, method ...int) interface{} {
 func calculateSkewType1(dl *insyra.DataList, highPrecision ...bool) interface{} {
 	n := new(big.Rat).SetFloat64(conv.ParseF64(dl.Len()))
 	nReciprocal := new(big.Rat).Inv(n)
-	m1 := dl.Mean(true).(*big.Rat)
+	m1 := new(big.Rat).SetFloat64(dl.Mean())
 	toM2Fn := func() *big.Rat {
 		var m2Cal = new(big.Rat)
 		for _, v := range dl.Data() {
diff --git a/stats/ttest.go b/stats/ttest.go
index f2109fa..6584d9f 100644
--- a/stats/ttest.go
+++ b/stats/ttest.go
@@ -26,7 +26,7 @@ func SingleSampleTTest(data insyra.IDataList, mu float64) *TTestResult {
 	}
 
 	// 計算樣本均值
-	mean := data.Mean(false).(float64)
+	mean := data.Mean()
 
 	// 計算標準差和標準誤差
 	stddev := data.Stdev(false).(float64)
@@ -56,8 +56,8 @@ func TwoSampleTTest(data1, data2 insyra.IDataList, equalVariance bool) *TTestRes
 	}
 
 	// 計算兩個樣本的均值
-	mean1 := data1.Mean(false).(float64)
-	mean2 := data2.Mean(false).(float64)
+	mean1 := data1.Mean()
+	mean2 := data2.Mean()
 
 	// 計算兩個樣本的標準差
 	stddev1 := data1.Stdev(false).(float64)
@@ -109,7 +109,7 @@ func PairedTTest(data1, data2 insyra.IDataList) *TTestResult {
 	}
 
 	// 計算差值的均值和標準差
-	meanDiff := insyra.NewDataList(diffs).Mean(false).(float64)
+	meanDiff := insyra.NewDataList(diffs).Mean()
 	stddevDiff := insyra.NewDataList(diffs).Stdev(false).(float64)
 
 	// 計算 t 值