EViews-style workflow
Goal
This page is the bridge for analysts moving an existing EViews time-series workflow into Python. StochX keeps the order of operations and familiar vocabulary while making each step explicit Python code.
The workfile
Workfile stores named series, an optional frequency, and the current estimation sample.
from stochx.timeseries import Workfile
wf = Workfile.from_dataframe(frame, frequency="M")
print(wf.info())
print(wf.names())
wf.set_sample(12, 120)
A CSV can be loaded directly:
wf = Workfile.from_csv(
"macro.csv",
date_column="DATE",
frequency="M",
)
Series expressions
StochX accepts EViews-inspired expressions for lags, leads and transformations:
| EViews-style expression | Meaning |
|---|---|
Y |
current series |
Y(-1) |
first lag |
Y(1) |
first lead |
D(Y) |
first difference |
DLOG(Y) |
log difference |
LOG(Y) |
natural logarithm |
@TREND |
deterministic trend term where supported |
wf.eval("GDP(-1)")
wf.eval("D(GDP)")
wf.generate("DGDP", "D(GDP)")
Equations
The familiar equation specification is kept as a string:
eq = wf.ls("GDP C CONS CONS(-1)", name="EQ01")
print(eq.summary())
print(eq.table())
print(eq.interpret())
C denotes the intercept in the EViews-style specification. Lags are written directly in the expression.
Regression with ARMA errors
Serial-correlation terms belong to the disturbance process rather than the regressors:
eq = wf.ls(
"TBILL C LOG(M1) LOG(CPI) LOG(IP) @TREND AR(1) MA(1)",
name="EQ20",
)
The result exposes coefficients, standard errors, t-values, p-values, fit statistics, roots and covariance information where supported.
Stationarity decision
The same workfile can feed DF/ADF procedures:
from stochx.timeseries import adf
result = adf(wf["GDP"], regression="c", lags=1, autolag=None)
print(result.summary())
print(result.interpret())
The point is not to hide the unit-root decision inside a model selector. The deterministic specification, lag order, statistic and critical values remain visible.
Mapping mindset
A useful migration pattern is:
| EViews habit | StochX habit |
|---|---|
| Workfile | Workfile |
| Series view | TimeSeries / Workfile[name] |
| Generate series | wf.generate(name, expression) |
| Lag | Y(-1) / wf.lag("Y") |
| Difference | D(Y) / wf.diff("Y") |
| Equation LS | wf.ls("Y C X") |
| ADF | adf(...) |
| Correlogram | correlogram(...) |
| ARMA/ARIMA | fit_arma, fit_arima, estimate, Box–Jenkins workflow |
| Forecast | forecast_* helpers |
What “compatible” means
StochX is EViews-inspired, not an attempt to reproduce the EViews application itself. It follows the same statistical workflow and vocabulary where that interface is implemented. Numerical parity is treated as a testable benchmark for specific frozen fixtures rather than as an implicit guarantee for every possible dataset.
Next steps
Continue with Data and series, then Transformations, Stationarity, and Correlation.