0
# Variational Inference Results
1

2
Container for Automatic Differentiation Variational Inference (ADVI) results and approximate posterior samples. The CmdStanVB class provides access to variational parameters and generated posterior samples.
3

4
## Capabilities
5

6
### Variational Parameters
7

8
Access the mean-field or full-rank variational parameters that define the approximate posterior.
9

10
```python { .api }
11
def variational_params_np(self):
12
    """
13
    Get variational parameters as NumPy array.
14
    
15
    Returns:
16
    np.ndarray: Variational parameter values (means and standard deviations)
17
    """
18

19
def variational_params_pd(self):
20
    """
21
    Get variational parameters as pandas DataFrame.
22
    
23
    Returns:
24
    pd.DataFrame: Parameters with names as index
25
    """
26

27
def variational_params_dict(self):
28
    """
29
    Get variational parameters as dictionary.
30
    
31
    Returns:
32
    dict: Mapping from parameter names to arrays
33
    """
34
```
35

36
### Posterior Samples
37

38
Access samples drawn from the variational approximation.
39

40
```python { .api }
41
def variational_sample(self):
42
    """
43
    Get posterior sample from variational approximation.
44
    
45
    Returns:
46
    np.ndarray: Draws from approximate posterior (draws, parameters)
47
    """
48

49
def variational_sample_pd(self):
50
    """
51
    Get posterior sample as pandas DataFrame.
52
    
53
    Returns:
54
    pd.DataFrame: Draws with parameter names as columns
55
    """
56
```
57

58
### Variable Access
59

60
```python { .api }
61
def stan_variable(self, var):
62
    """
63
    Get posterior sample for specific Stan variable.
64
    
65
    Parameters:
66
    - var (str): Variable name
67
    
68
    Returns:
69
    np.ndarray: Draws for the variable
70
    """
71

72
def stan_variables(self):
73
    """
74
    Get all Stan variables as dictionary.
75
    
76
    Returns:
77
    dict: Mapping from variable names to draw arrays
78
    """
79
```
80

81
### File Operations
82

83
```python { .api }
84
def save_csvfiles(self, dir=None):
85
    """
86
    Save CSV output files to directory.
87
    
88
    Parameters:
89
    - dir (str or PathLike, optional): Target directory
90
    
91
    Returns:
92
    None
93
    """
94
```
95

96
## Properties
97

98
```python { .api }
99
# Variational information
100
vb.columns              # int: Number of parameters
101
vb.column_names         # Tuple[str, ...]: Parameter names
102
vb.eta                  # float: Final step size
103
vb.metadata            # InferenceMetadata: Run configuration and timing
104
```
105

106
## Usage Examples
107

108
```python
109
# Run variational inference
110
vb = model.variational(data=data, algorithm="meanfield", draws=2000)
111

112
# Access variational parameters
113
var_params = vb.variational_params_dict()
114
print("Variational means:", var_params)
115

116
# Get posterior samples
117
posterior_samples = vb.variational_sample()
118
theta_draws = vb.stan_variable("theta")
119

120
print(f"Posterior mean for theta: {theta_draws.mean()}")
121
print(f"Posterior std for theta: {theta_draws.std()}")
122
```