Autocorrelation function is invariant in time, i.e. correlation between error e(t) and e(t-k) only depends on k not t.

yes

no

Autocorrelation always decays exponentially with k between errors e(t) and e(t-k)

yes

no

In a linear model y=x*beta+e, with T observation, k bona fide exogenous variables, and n lagged dependent variables, the covariance matrix of errors has dimensions

1. Tx(k+n+1)
2. Txk
3. TxT
4. (k+1)x(k+1)

Autocorrelation function r(k) always depends only on the distance between errors e(t) and e(t-k), not the time t.

yes

no

Durbin Watson test statistic is from chi-square distribution

yes

no

Weakly stationary condition is impossible to prove or test, it's a pure theoretical concept

yes

no

Unless both y and x are stationary, OLS cannot be applied to a linear model y=x*beta+e

yes

no

ARIMAX and regARIMA are synonyms

yes

no

Apply backshift operator (1-2B+4B^2) to x(t-2)

1. x(t-2)-2 x(t-2)+4 [x(t-2)]^2
2. x(t-2)-2 x(t-3)+4 x(t-4)
3. x(t-2)-2 x(t-2)+4 x(t-2)
4. x(t)-2 x(t-1)+4 x(t-2)

Regression with ARIMA errors is an inferior model to ARIMAX

yes

no

GARCH is a stochastic volatility model because it has a stochastic term r(t-1)^2, a square of return

yes

no

Simple moving average is an example of low pass filter

Yes

No

Update parameter lambda in EWMA volatility metric cannot be estimated, it's arbitrarily chosen by an expert based on experience or industry practice

Yes

No

Pick the components of GARCH estimate of volatility v(t)

1. squared return r(t)
2. Long run variance
3. previous estimate of volatility v(t-1)
4. square return r(t-1)

Chi-square goodness of fit test is only applicable to discrete probability distributions

Yes

No

Binomial goodness of fit tests are for only for VaR Monet Carlo simulations where the underlying distributions are binomial

Yes

No