Pull Data from Prediction Signal in Workbench to Data Lab

[Rand]

Is there any way to get the formula, y_intercept, and the coefficient values of the predicted signal in Data Lab? 

[Selmane]

Hello Rand,

yes, you can use the regressionModelOLS() Formula. Below is a screenshot of the formula help:

You can use this formula in Data Lab directly using spy library.

Thanks,

--Selmane

[Thorsten Vogt]

Hi Rand, 

if you want to get the data for an already existing item in SDL you can use the following code. 

import seeq.sdk as sdk

api = sdk.FormulasApi(spy.client)
item = spy.search({"ID": "8A90CF9E-4E83-4753-87E8-D76B52ABC27C"}, all_properties=True)
formula = item.iloc[0]["Formula"].split('.predict')[0]
parameters=item.iloc[0]["Formula Parameters"]

formula_result = api.run_formula(formula=formula, parameters=parameters)
print(formula_result)

The result is a JSON structure containing all the relevant information:

{'capsules': None,
 'errors': [],
 'metadata': {},
 'regression_output': {'adjusted_r_squared': 0.38955801847929894,
                       'error_sum_squares': 3.4469623889868926,
                       'intercept': 0.24317408479080208,
                       'intercept_standard_error': 1.9738767365485246,
                       'is_uncertain': False,
                       'r_squared': 0.440428183606024,
                       'regression_sum_squares': 2.7130376110131085,
                       'suggested_p_value_cutoff': 0.19152956820564848},
 'return_type': 'RegressionModel',
 'samples': None,
 'scalar': None,
 'status_message': None,
 'table': {'data': [[-0.05715138991243394,
                     0.042461052848400904,
                     0.1914295682056485],
                    [0.5433690783000202,
                     0.959107303992974,
                     0.5765165220597079]],
           'headers': [{'name': 'coefficient', 'type': 'number', 'units': ''},
                       {'name': 'coefficientStandardError',
                        'type': 'number',
                        'units': ''},
                       {'name': 'pValue', 'type': 'number', 'units': ''}]},
 'upgrade_details': None,
 'warning_count': None,
 'warning_logs': None}

The table.data array contains the coefficient, coefficientStandardError and p-Value for the paramers, which can be extracted with some additional code. Depending on the regression method you use you may need to adjust it. 

import re
import pandas as pd

# Regular expression to capture the parameters of regressionModelOLS
match = re.search(r'regressionModelOLS\((.*)\)', formula)

if match:
    # Split the parameters by commas at the top level (ignoring nested commas)
    params = re.split(r',\s*(?![^()]*\))', match.group(1))
    # Get all items after the 3rd parameter
    index_cols = params[2:]
    
headers = []
for header in formula_result.table.headers:
    headers.append(header.name)

# Create the DataFrame
df = pd.DataFrame(formula_result.table.data, index=index_cols, columns=headers)
df

Regards,

Thorsten