nardl_auto_case {ardl.nardl} | R Documentation |

This function finds the best NARDL model specification and conduct bounds test by relying on the general to specific approach.

```
nardl_auto_case(x, decomp, dep_var, control = NULL, c_q_order = c(2),
p_order = c(3), q_order, gets_pval = 0.1,
order_l = order_l, graph_save = FALSE)
```

`x` |
Dataframe |

`decomp` |
A character vector. The variable to be decomposed to positive (pos) and negative (neg) variable. |

`dep_var` |
A character vector. The dependent variable |

`control` |
A character vector. Default is NULL. The second dependent variable. |

`c_q_order` |
Integer. Maximum number of lags for 'control' |

`p_order` |
An integer. Lag differenced adopted for the differenced response variable |

`q_order` |
An integer. Lag differenced adopted for the differenced explanatory variable(s) |

`gets_pval` |
The p- value which served as criteria for eliminating non-significant variable in the course of obtaining the best model based on the Schwarz information criteria. |

`order_l` |
Integer. Needed for the autocorrelation and heteroscedasticity test |

`graph_save` |
Logical. If TRUE, displays the stability plots |

The procedure of the general-to-specific approach in obtaining the parsimonious model involves conducting the multi-path backwards elimination; tests both single and multiple hypothesis tests, diagnostics tests and goodness-of-fit measures. See page 5 - 6 of Sucarrat (2021) for more details.

The value for gets_pval is influential the final model based on the multipath backward elimination. For more details on the general-to-specific approach, see the vignette of the 'gets' package.

`Parsimonious_NARDL_fit ` |
Return an estimated general-to-specific NARDL model. |

`Parsimonious_ECM_fit ` |
Return an estimated general-to-specific error correction model. |

`Summary_uecm_fit ` |
Return the summary of 'Parsimonious_ECM_fit' |

`ecm_diagnostics_test ` |
Return the diagnostic test for the 'Parsimonious_ECM_fit'. The diagnostic tests indicate the Breusch-Godfrey test for higher-order serial correlation (BG_SC_lm_test). The Engle (1982) test for conditional heteroscedasticity (LM_ARCH_test). The test for non-normality is that of Jarque and Bera (1980). The RESET null hypothesis adopted implies - including the 2nd - degree terms improve the fit (over the model specified). |

`longrun_asym ` |
Return the estimated longrun asymmetric test |

`Shortrun_asym ` |
Return the estimated short-run asymmetric test.If one of the decomposed variable does not appear among the shortrun differenced variables of the parsimonious model, The value returned is a wald test of whether the sum of the coefficients of the remaining decomposed variable included does not have any significant effect on the best model |

`cointegration ` |
Return the F statistic, the upper and lower critical values for PSS (2001) bounds test. Please, disregard the tstat on the cointegration test. |

`Longrun_relation ` |
The longrun relation |

Do not differenced the variables to be adopted in this function and all other functions for NARDL and ARDL estimation. The package inherently takes the difference and produced output with a prefix (D.) to the variable name and suffix the variable name with underscore (_) and the lag value.

Sucarrat, G. User-Specified General-to-Specific (GETS) and Indicator Saturation (ISAT) Methods. 28th September 2021. https://mirror.epn.edu.ec/CRAN/web/packages/gets/vignettes/user-defined-gets-and-isat.pdf

`gets`

`gets_nardl_uecm`

`ardl_uecm`

`auto_case_ardl`

```
## Not run:
data("fuel_price")
out1 <- nardl_auto_case(x = fuel_price,
decomp = 'wti',
dep_var = 'fpp',
control = 'bdc',
c_q_order = c(5),
p_order = c(5),
q_order = c(6),
gets_pval = 0.1,
order_l = 4,
graph_save = FALSE)
out1
out2 <- nardl_auto_case(x = fuel_price,
decomp = 'wti',
dep_var = 'fpp',
control = NULL,
c_q_order = c(4),
p_order = c(5),
q_order = c(6),
gets_pval = 0.02,
order_l = 4,
graph_save = FALSE)
out2
## End(Not run)
```

[Package *ardl.nardl* version 1.2.3 Index]