Update app.py

app.py

@@ -4,14 +4,13 @@ from scipy import stats
 from typing import List, Dict, Any, Union, Tuple
 import json
 
-def independent_t_test(group1: str, group2: str, equal_var: bool = True, alternative: str = "two-sided") -> Dict[str, Any]:
+def student_t_test(group1: str, group2: str, alternative: str = "two-sided") -> Dict[str, Any]:
     """
-    Perform an independent samples t-test between two groups.
+    Perform Student's t-test (assumes equal variances) between two independent groups.
     
     Args:
         group1 (str): Comma-separated values for group 1 (e.g., "1.2,2.3,3.4,2.1")
         group2 (str): Comma-separated values for group 2 (e.g., "2.1,3.2,4.1,3.5")
-        equal_var (bool): If True, perform standard t-test assuming equal variances. If False, perform Welch's t-test
         alternative (str): Alternative hypothesis - 'two-sided', 'less', or 'greater'
     
     Returns:
@@ -25,24 +24,76 @@ def independent_t_test(group1: str, group2: str, equal_var: bool = True, alterna
         if len(data1) < 2 or len(data2) < 2:
             return {"error": "Each group must have at least 2 observations"}
         
-        # Perform t-test
-        t_stat, p_value = stats.ttest_ind(data1, data2, equal_var=equal_var, alternative=alternative)
+        # Perform Student's t-test (equal variances)
+        t_stat, p_value = stats.ttest_ind(data1, data2, equal_var=True, alternative=alternative)
         
         # Calculate descriptive statistics
         desc1 = {"mean": np.mean(data1), "std": np.std(data1, ddof=1), "n": len(data1)}
         desc2 = {"mean": np.mean(data2), "std": np.std(data2, ddof=1), "n": len(data2)}
         
-        # Degrees of freedom
-        if equal_var:
-            df = len(data1) + len(data2) - 2
-        else:
-            # Welch's formula for unequal variances
-            s1_sq, s2_sq = desc1["std"]**2, desc2["std"]**2
-            n1, n2 = desc1["n"], desc2["n"]
-            df = (s1_sq/n1 + s2_sq/n2)**2 / ((s1_sq/n1)**2/(n1-1) + (s2_sq/n2)**2/(n2-1))
+        # Degrees of freedom (pooled)
+        df = len(data1) + len(data2) - 2
         
-        # Effect size (Cohen's d)
-        pooled_std = np.sqrt(((len(data1)-1)*desc1["std"]**2 + (len(data2)-1)*desc2["std"]**2) / (len(data1)+len(data2)-2))
+        # Effect size (Cohen's d using pooled standard deviation)
+        pooled_std = np.sqrt(((len(data1)-1)*desc1["std"]**2 + (len(data2)-1)*desc2["std"]**2) / df)
+        cohens_d = (desc1["mean"] - desc2["mean"]) / pooled_std
+        
+        # Interpretation
+        significance = "significant" if p_value < 0.05 else "not significant"
+        effect_size_interp = "small" if abs(cohens_d) < 0.5 else "medium" if abs(cohens_d) < 0.8 else "large"
+        
+        return {
+            "test_type": "Student's t-test (equal variances assumed)",
+            "t_statistic": round(t_stat, 4),
+            "p_value": round(p_value, 6),
+            "degrees_of_freedom": df,
+            "cohens_d": round(cohens_d, 4),
+            "pooled_std": round(pooled_std, 4),
+            "group1_stats": desc1,
+            "group2_stats": desc2,
+            "result": f"The difference between groups is {significance} (p = {p_value:.6f})",
+            "effect_size": f"Effect size (Cohen's d = {cohens_d:.4f}) is {effect_size_interp}",
+            "alternative_hypothesis": alternative,
+            "assumption": "Assumes equal population variances"
+        }
+    except Exception as e:
+        return {"error": f"Error performing Student's t-test: {str(e)}"}
+
+def welch_t_test(group1: str, group2: str, alternative: str = "two-sided") -> Dict[str, Any]:
+    """
+    Perform Welch's t-test (does not assume equal variances) between two independent groups.
+    
+    Args:
+        group1 (str): Comma-separated values for group 1 (e.g., "1.2,2.3,3.4,2.1")
+        group2 (str): Comma-separated values for group 2 (e.g., "2.1,3.2,4.1,3.5")
+        alternative (str): Alternative hypothesis - 'two-sided', 'less', or 'greater'
+    
+    Returns:
+        dict: Test results including t-statistic, p-value, degrees of freedom, and interpretation
+    """
+    try:
+        # Parse input data
+        data1 = [float(x.strip()) for x in group1.split(',') if x.strip()]
+        data2 = [float(x.strip()) for x in group2.split(',') if x.strip()]
+        
+        if len(data1) < 2 or len(data2) < 2:
+            return {"error": "Each group must have at least 2 observations"}
+        
+        # Perform Welch's t-test (unequal variances)
+        t_stat, p_value = stats.ttest_ind(data1, data2, equal_var=False, alternative=alternative)
+        
+        # Calculate descriptive statistics
+        desc1 = {"mean": np.mean(data1), "std": np.std(data1, ddof=1), "n": len(data1)}
+        desc2 = {"mean": np.mean(data2), "std": np.std(data2, ddof=1), "n": len(data2)}
+        
+        # Welch's degrees of freedom formula
+        s1_sq, s2_sq = desc1["std"]**2, desc2["std"]**2
+        n1, n2 = desc1["n"], desc2["n"]
+        df = (s1_sq/n1 + s2_sq/n2)**2 / ((s1_sq/n1)**2/(n1-1) + (s2_sq/n2)**2/(n2-1))
+        
+        # Effect size (Cohen's d using separate standard deviations)
+        # For unequal variances, use the average of the two standard deviations
+        pooled_std = np.sqrt((desc1["std"]**2 + desc2["std"]**2) / 2)
         cohens_d = (desc1["mean"] - desc2["mean"]) / pooled_std
         
         # Interpretation
@@ -50,7 +101,7 @@ def independent_t_test(group1: str, group2: str, equal_var: bool = True, alterna
         effect_size_interp = "small" if abs(cohens_d) < 0.5 else "medium" if abs(cohens_d) < 0.8 else "large"
         
         return {
-            "test_type": f"Independent t-test ({'equal variances' if equal_var else 'unequal variances'})",
+            "test_type": "Welch's t-test (unequal variances)",
             "t_statistic": round(t_stat, 4),
             "p_value": round(p_value, 6),
             "degrees_of_freedom": round(df, 2),
@@ -59,10 +110,11 @@ def independent_t_test(group1: str, group2: str, equal_var: bool = True, alterna
             "group2_stats": desc2,
             "result": f"The difference between groups is {significance} (p = {p_value:.6f})",
             "effect_size": f"Effect size (Cohen's d = {cohens_d:.4f}) is {effect_size_interp}",
-            "alternative_hypothesis": alternative
+            "alternative_hypothesis": alternative,
+            "assumption": "Does not assume equal population variances (recommended default)"
         }
     except Exception as e:
-        return {"error": f"Error performing t-test: {str(e)}"}
+        return {"error": f"Error performing Welch's t-test: {str(e)}"}
 
 def paired_t_test(before: str, after: str, alternative: str = "two-sided") -> Dict[str, Any]:
     """
@@ -364,16 +416,26 @@ def correlation_test(x_values: str, y_values: str, method: str = "pearson") -> D
 demo = gr.TabbedInterface(
     [
         gr.Interface(
-            fn=independent_t_test,
+            fn=student_t_test,
+            inputs=[
+                gr.Textbox(placeholder="1.2,2.3,3.4,2.1", label="Group 1 (comma-separated)"),
+                gr.Textbox(placeholder="2.1,3.2,4.1,3.5", label="Group 2 (comma-separated)"),
+                gr.Dropdown(["two-sided", "less", "greater"], value="two-sided", label="Alternative hypothesis")
+            ],
+            outputs=gr.JSON(),
+            title="Student's T-Test",
+            description="Compare means between two independent groups (assumes equal variances)"
+        ),
+        gr.Interface(
+            fn=welch_t_test,
             inputs=[
                 gr.Textbox(placeholder="1.2,2.3,3.4,2.1", label="Group 1 (comma-separated)"),
                 gr.Textbox(placeholder="2.1,3.2,4.1,3.5", label="Group 2 (comma-separated)"),
-                gr.Checkbox(value=True, label="Equal variances"),
                 gr.Dropdown(["two-sided", "less", "greater"], value="two-sided", label="Alternative hypothesis")
             ],
             outputs=gr.JSON(),
-            title="Independent T-Test",
-            description="Compare means between two independent groups"
+            title="Welch's T-Test",
+            description="Compare means between two independent groups (does not assume equal variances - recommended)"
         ),
         gr.Interface(
             fn=paired_t_test,
@@ -432,9 +494,9 @@ demo = gr.TabbedInterface(
             description="Test correlation between two variables"
         )
     ],
-    tab_names=["Independent T-Test", "Paired T-Test", "One-Sample T-Test", "ANOVA", "Chi-Square", "Correlation"]
+    tab_names=["Student's T-Test", "Welch's T-Test", "Paired T-Test", "One-Sample T-Test", "ANOVA", "Chi-Square", "Correlation"]
 )
-
+        
 if __name__ == "__main__":
     print(f"Gradio version: {gr.__version__}")
     demo.launch(mcp_server=True)