Actually, I'd say just the opposite. Multicolinearity is often scoffed at as a concern. The only time this is a real issue is when one variable can be written as an exact linear function of others in the model (a male dummy variable would be exactly equal to a constant/intercept term minus a female dummy variable; hence, you can't have all three in your model). A prime example is Goldberger's comparison to "micronumerousity."

Perfect multicolinearity means that your model cannot be estimated; (not perfect) multicolinearity often leads to large standard errors, but no bias or real problems; heteroskedasticity means that your standard errors are incorrect and your estimates are inefficient.

First, I would create a model that yields the parameter estimates as I want to interpret them (level change, percent change, etc.) by using logs as appropriate. Then, I would test for heteroskedasticity. The most accepted option is to simply use robust standard errors to give you correct standard errors, but for inefficient parameter estimates. Alternatively, you can use weighted least squares to get efficient estimates, but this has become less common unless you know the relationship between the variances of your observations (they each depend upon the size of the observation---like population of a country). Indeed, in cross section econometrics using a data set of any real size, robust standard errors have become required irrespective of the outcome of a BP test; they are applied almost automatically.

There isn't a good test for endogeneity. You're real problem is that the regressor is correlated with the *error*; OLS will *force* the regressor to be uncorrelated with the *residual*. So you won't find any correlation there. Endogeneity is what makes econometrics hard and is a whole topic unto itself.

*How do I select the best ARIMA model (by trying all different orders and checking the best MASE/MAPE/MSE? where the selection of performance measurement can be a discussion in it's own..)*

Out of sample risk estimates are the gold standard for performance evaluation, and therefore for model selection. Ideally, you cross-validate so that your risk estimates are averaged over more data. FPP explains one cross-validation method for time series. See Tashman for a review of other methods:

Tashman, L. J. (2000). Out-of-sample tests of forecasting accuracy: an analysis and review. International Journal of Forecasting, 16(4), 437–450. doi:10.1016/S0169-2070(00)00065-0

Of course, cross-validation is time consuming and so people often resort to using in-sample criteria to select a model, such as AIC, which is how auto.arima selects the best model. This approach is perfectly valid, if perhaps not as optimal.

*If I generate a new model and forecast for every new day forecast (as in online forecasting), do I need to take the yearly trend into account and how? (as in such a small subset my guess would be that the trend is neglible)*

I'm not sure what you mean by yearly trend. Assuming you mean yearly seasonality, there's not really any way to take it into account with less than a year's worth of data.

*Would you expect that the model order stays the same throughout the dataset, i.e. when taking another subset will that give me the same model?*

I would expect that barring some change to how the data are generated, the most correct underlying model will be the same throughout the dataset. However, that's not the same as saying that the model selected by any procedure (such as the procedure used by auto.arima) will be the same if that procedure is applied to different subsets of the data. This is because the variability due to sampling will result in variability in the results of the model selection procedure.

*What is a good way, within this method to cope with holidays? Or is ARIMAX with external holiday dummies needed for this?*

External holiday dummies is the best approach.

*Do I need to use Fourier series approach to try models with *`seasonality=672`

as discussed in Long seasonal periods?

You need to do something, because as mentioned in that article, the arima function in R does not support seasonal periods greater than 350. I've had reasonable success with the Fourier approach. Other options include forecasting after seasonal decomposition (also covered in FPP), and exponential smoothing models such as bats and tbats.

*If so would this be like *`fit<-Arima(timeseries,order=c(0,1,4), xreg=fourier(1:n,4,672)`

(where the function fourier is as defined in Hyndman's blog post)

That looks correct. You should experiment with different numbers of terms. Note that there is now a `fourier`

function in the forecast package with a slightly different specification that I assume supersedes the one on Hyndman's blog. See the help file for syntax.

*Are initial P and Q components included with the fourier series?*

I'm not sure what you're asking here. P and Q usually refer to the degrees of the AR and MA seasonal components. Using the fourier approach, there are no seasonal components and instead there are covariates for fourier terms related to season. It's no longer seasonal ARIMA, it's ARIMAX where the covariates approximate the season.

## Best Answer

Your $p$-value is very small, so you actually will reject the null hypothesis at any sensible significance level.

(I just wonder if you are using the test correctly. Are you supplying any explanatory variables for the test? Why do you have zero degrees of freedom there:

`df=0`

?)However, BP test does not test for ARCH patterns. It tests for another kind of heteroskedasticity: that the magnitude of errors (measured by squared errors) varies with the explanatory variables in the model (see Wikipedia for more details).

Meanwhile, ARCH is autoregressive behaviour in squares of errors, which is quite different from the above. Therefore, ARCH structure in model errors is not an implication of the test result.