results.py¶
pycmtensor.results
¶
PyCMTensor results module
This module provides the results of the estimated model and output formatting
Results()
¶
Bases: object
Base class object to hold model results and scores
Attributes:
| Name | Type | Description |
|---|---|---|
build_time |
str
|
string formatted time stamp of the duration of the build stage |
train_time |
str
|
string formatted time stamp of the duration of the training stage |
epochs_per_sec |
float
|
number of epochs of the training dataset per second, benchmark for calculation speed |
n_params |
int
|
total number of parameters, used for calculating statistics |
n_train |
int
|
number of training samples used |
n_valid |
int
|
number of validation samples used |
seed |
int
|
random seed value used |
null_loglikelihood |
float
|
null log likelihood of the model |
best_loglikelihood |
float
|
the final estimated model log likelihood |
best_valid_error |
float
|
final estimated model validation error |
best_epoch |
int
|
the epoch number at the final estimated model |
gnorm |
float
|
the gradient norm at the final estimated model |
hessian_matrix |
ndarray
|
the 2-D hessian matrix |
bhhh_matrix |
ndarray
|
the 3-D bhhh matrix where the 1st dimension is the length of the dataset and the last 2 dimensions are the matrix for each data observation |
statistics_graph |
dict
|
a dictionary containing the learning_rate, training, and validation statistics |
betas |
dict
|
a dictionary containing the Beta coefficients |
params |
dict
|
a dictionary containing all model coefficients |
rho_square()
¶
rho square value of the model
Returns:
| Type | Description |
|---|---|
float
|
analogue for the model fit |
rho_square_bar()
¶
McFadden's adjusted rho square
Returns:
| Type | Description |
|---|---|
float
|
the adjusted McFadden's rho square value |
loglikelihood_ratio_test()
¶
Log likelihood ratio test
Returns:
| Type | Description |
|---|---|
float
|
the log likelihood ratio test: $-2 times (NLL-LL) |
AIC()
¶
Akaike information criterion
Returns:
| Type | Description |
|---|---|
float
|
the AIC of the model |
BIC()
¶
Bayesian information criterion, adjusted for the number of parameters and number of training samples
Returns:
| Type | Description |
|---|---|
float
|
the BIC of the model |
benchmark()
¶
benchmark statistics
Returns:
| Type | Description |
|---|---|
DataFrame
|
Summary of the model performance benchmark |
model_statistics()
¶
model statistics
Returns:
| Type | Description |
|---|---|
DataFrame
|
Summary of the model statistics |
beta_statistics()
¶
Beta statistics
Returns:
| Type | Description |
|---|---|
DataFrame
|
Summary of the estimated Beta statistics |
model_correlation_matrix()
¶
Correlation matrix calculated from the hessian
Returns:
| Type | Description |
|---|---|
DataFrame
|
The correlation matrix |
model_robust_correlation_matrix()
¶
Robust correlation matrix calculated from the hessian and bhhh
Returns:
| Type | Description |
|---|---|
DataFrame
|
The robust correlation matrix |
show_training_plot(sample_intervals=1)
¶
Displays the statistics graph as a line plot
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
sample_intervals |
int
|
plot only the n-th point from the statistics |
1
|
Returns:
| Type | Description |
|---|---|
DataFrame
|
The statistics as a pandas dataframe |