We’re rewarding the question askers & reputations are being recalculated! Read more.

Hot answers tagged

10

The whites.htest() function implements White's test for heteroskedasticity for vector autoregressions (VAR). It requires a varest object as input. However, from your description it seems that your model is not a VAR (vector autoregression) but a simple linear model. Hence, the model should be estimated by lm() as previously suggested in the comments. Then ...


5

As I read the results, you have two cointegrating equations, or two cointegrating vectors. This translates into cointegration rank being equal to one (number of variables in the system minus the number of cointegrating vectors: $3-2=1$). Two cointegrating vectors is not the same as cointegration order being equal to two. Your cointegration order is equal to ...


5

The formulation of an ARIMA model with exogenous regressors is not generally the same as a linear regression model with lagged dependent variables. To my knowledge, the formulation in software packages for the ARIMA model with exogenous regressors is the following: $$ \left[ y(t) - \beta_0 - \beta_3 \hbox{levelshift}(t) \right] = \beta_1 \left[y(t-1) - \...


4

VAR models are routinely used with seasonal data, e.g. in macroeconomics where most of the time series (such as GDP or unemployment) are seasonal. Seasonality is handled either (1) outside of the model (by seasonally adjusting the series before fitting a VAR model) or (2) within the model (by including seasonal dummy variables, for example). For (1), ...


4

The behavior that you see is due to the presample variance option in EViews. rugarch uses the variance of all data points and EViews uses backcasting using a parameter of 0.7. More precisely, EViews uses this formula for initialization of the variance: $\sigma_0^2 = \lambda \hat\sigma^2 + (1-\lambda) \sum_{t=0}^T \lambda^{T-t-1} \cdot \varepsilon_{T-t}^2$ ...


3

I'll answer your questions pertaining to cointegration. 1) If the context of your exercise is the forecasting of a particular dependent variable by using a set of independent variables as opposed to jointly forecasting a set of variables, then you want to explore an ECM not a VECM, with the latter one being potentially overkill. 2) In your process of ...


3

Regarding significance: That is partly dependent on sample size. I googled and the NLYS data set is pretty big (N ~ 10,000) so even small effects will be statistically significant. Look at effect sizes. Regarding ethblack and ethwhite: How did the data set code race? You are here comparing each of these groups to anyone who was in neither group. And, as ...


3

In general you want to adjust for confounders - things that you believe affect both your outcome (earnings) and your exposure (race). Adjusting for variables downstream from race means that you are eliminating the ability to detect an effect of race on earnings operating through those variables. For instance, maybe there is an effect of race on earnings ...


2

I think the reason why the p-values are not reported is because the Q-statistic is follows a Chi-squared distribution, where the d.f. = the number of lags (e.g. at lag 2, d.f. = 2). However, for the residuals calculated from an ARMA or ARIMA estimation, the d.f. should be adjusted for the number of ARMA terms. So I suppose the output above comes from an ARMA ...


1

1) In ADF, the $H_0$ is the presence of unit root and $H_a$ is for stationary (weakly dependent) series. In your both examples, you don't reject $H_0$. To reject $H_0$, you want the "Augmented Dickey-Fuller test statistic" to be farther from zero (more negative) than the critical value at a significance level chosen. To find out the order of integration, you ...


1

If the only problem is that the stationary variables will get differenced and you want to avoid that, then there is a simple hack: supply cumulative sums of those variables (i.e. "integrate" them). When the cumulative sums get differenced, they yield the original variables as differencing undoes cumulative summation. In R you can do cumulative summation via ...


1

why does exchange rate (LNEX) have a coefficient of 0? Because cointegration vector is statistically restricted (or normalized) in this way, in order to guarantee its identifiability. If you have theoretical restrictions, you can ignore this type of restrictions and use them. In other word, consider the following cointegrated VAR model: $$ \Delta\mathbf{y}...


1

From my perspective you have already picked data that tends to be not normal as you are focussing on somewhat extreme cases. How do the residuals look if you include the number of earthquakes of all magnitudes? Additionally you could think about control variables that capture some of the structure in the residuals. Another thing I would recommend you to do ...


1

Off topic (and I do not know the answer). If the appropriate lags for $y$ and $x$ are different, $p \neq q$, forcing a common lag $r$ will be suboptimal. If $r \geqslant \max(p,q)$, there will be unnecessarily many parameters in the model resulting in increased estimation variance and loss of power in the Granger causality test. If $r \leqslant \min(p,q)$, ...


1

You are rejecting the null at 5% sig level, not stationary.


1

To get a better estimation on the income elasticity of demand you should include all the X-variables in the model. The estimated coefficient you will get for your income elasticity variable in the complete model will be considered "Income elasticity of demand while holding other X-variables constant". If you omit the other X-variables, you do indeed assume ...


1

Seasonal dummies and seasonal auto-regressive structure are often "competitors" when finding the best model. Your significant lag1 and lag2 effects often enable a clearer picture as to which approach is best i.e. seasonal dummies or seasonal memory (ARIMA) . This is why we recommend and implement a tournament approach to identifying an initial model. Care ...


1

First look at the plot of your time series data to have an idea about break point. If there seems to be a single structural break (also require literature support), you can test it by using 'Chow breakpoint test' and if there seems to be multiple structural breaks (also require literature support), you can test the exact breakpoint by using 'Bai-Perron ...


1

Don't know why Eviews does this (on R and Stata, those p-values are automatically computed, as one would expect). For lags 1 to 4, the AC's and PAC's are all within the 'error bars' implying that the p-values for those lags are all below 0.05. Therefore, residuals are white noise (are more accurately, you can't reject the null hypothesis that they are).


1

The coefficients are of comparable magnitudes in both models. While I do not know the actual implementations in R and Eviews, I am pretty sure that both implementations numerically maximize the (log-)likelihood function. The difference between the results may lay in different convergence criteria for numerical optimization algortihms used in the softwares. ...


1

This is not really an answer, but I hope it will be helpful. First, your "initial specification" is rather problematic. For the model to be estimable by OLS you would need the regressors (the right-hand-side variables) to be exogenous with respect to the regressand (the left-hand-side variable). Here you have interest rates that are anything but exogenous ...


1

Estimating a model in differences is one way to proceed. A unit root doesn't itself indicate cointegration is possible. How it works is this; If your dependent and independent variables share a common stochastic trend, they may cointegrate. Two series cointegrate in case of a genuine long run relationship between the two, i.e. they are both determined by ...


1

You conducted a Augmented Dickey Fuller test. The hypothesis of this test are $H_0$: "Process has unit root" vs. $H_1$: "Process has no unit root". The test statistic is $-37.22113$. Now you need to compare this with the critical values under $H_0$. The critical values are given with: $ 1\%: -3.435299 \\ 5\%: -2.863613 \\ 10\%: -2.567923.$ Since your ...


1

Once you have an estimated model, the standard errors and the t-values characterize the model coefficients but not the regressors. In a VECM the standard error associated with the error correction term (ECT) is the standard error of the corresponding estimated coefficient (the loading coefficient); it does not concern the ECT itself. Estimating a VECM is ...


1

There is no well-established technical criteria for what you are asking. But an answer to your question can be formulated. Before using your prior beliefs (which can be perfectly legitimate and useful), contemplate what the different specifications are giving you: When you include regressors that could account for non-stationarity (like individual ...


1

Since you didn't specify which version of Eviews you are using, I assume that you are using latest version (7). In that case, if you have a panel data in an Excel file, you just have to import that as foreign file. Please see here for details. I am also not sure whether there is any difference between control variable and independent variable. I think all ...


1

I am not sure if this is what you are asking, but one might conceive of an integrated process that looks something like this: $$y_{t} = y_{t-1} + \varepsilon$$ Feel free to define the distribution of $\varepsilon$ according to your needs and whimsy, and you have got yourself a nice little nonstationary random walk process. The "unit root" in the above ...


1

For the interpretation of Eviews output, just focus on top part. The lower one shows how the Eviews runs the regression. Eviews runs the regression in first difference form, so the null is coefficient on LOG_AUSTRIA(-1) is zero which means that there is an unit root.The alternate hypothesis is that it is less than zero, i.e., there is no unit root. Your ...


Only top voted, non community-wiki answers of a minimum length are eligible