We introduce a new statistic written as a sum of certain ratios of second order increments of partial sums process $S_n = \sum_{t=1}^n X_t$ of observations, which we call the Increment Ratio (IR) statistic. The IR statistic can be used for testing nonparametric hypotheses for $d-$integrated ($-1/2 < d < 3/2$) behavior of time series $X_t$, including short memory ($d=0$), (stationary) long-memory $(0 < d < 1/2)$ and unit roots ($d=1$). If $S_n$ behaves asymptotically as an (integrated) fractional Brownian motion with parameter $H =d + 1/2$, the IR statistic converges to a monotone function $\Lambda (d)$ of $d \in (-1/2,3/2)$ as both the sample size $N$ and the window parameter $m$ increase so that $N/m \to \infty$. For Gaussian observations $X_t$, we obtain a rate of decay of the bias $\mathbb{E} I\!R - \Lambda (d)$ and a central limit theorem $(N/m)^{1/2}(I\!R - \mathbb{E} I\!R) \rightarrow {\cal N}(0, \sigma^2(d))$, in the region $-1/2 < d < 5/4$. Graphs of the functions $\Lambda (d)$ and $\sigma (d)$ are included. A simulation study shows that the IR test for short memory ($d=0$) against stationary long-memory alternatives $(0 < d < 1/2)$ has good size and power properties and is robust against changes in mean, slowly varying trends and nonstationarities. We apply this statistic to sequences of squares of returns on financial assets and obtain a nuanced picture of the presence of long-memory in asset price volatility.
We consider the multiple change-point problem for multivariate time series,
including strongly dependent processes, with an unknown number of
change-points. We assume that the covariance structure of the series changes
abruptly at some unknown common change-point times. The proposed adaptive
method is able to detect changes in multivariate i.i.d., weakly and strongly
dependent series. This adaptive method outperforms the Schwarz criteria,
mainly for the case of weakly dependent data. We consider applications to
multivariate series of daily stock indices returns and series generated by an
artificial financial market.
Key words: adaptive methods, multivariate time series,
change-point detection, heteroskedasticity.
This chapter considers the multiple change-point problem for time series, including strongly dependent processes, with an unknown number of change-points. We propose an adaptive method for finding the segmentation, i.e., the sequence of change-points \(\boldsymbol{\tau}\) with the optimal level of resolution. This optimal segmentation \(\hat{\boldsymbol{\tau}}\) is obtained by minimizing a penalized contrast function \(J(\boldsymbol{\tau},\boldsymbol{y}) + \beta\; {\rm pen}(\boldsymbol{\tau})\). For a given contrast function \(J(\boldsymbol{\tau},\boldsymbol{y})\) and a given penalty function \({\rm pen}(\boldsymbol{\tau})\), the adaptive procedure for automatically choosing the penalization parameter \(\beta\) is such that the segmentation \(\hat{\boldsymbol{\tau}}\) does not strongly depend on \(\beta\). This algorithm is applied to the problem of detection of change-points in the volatility of financial time series, and compared with Vostrikova's (1981) binary segmentation procedure.
We present and study the performance of the semiparametric wavelet estimator for the long–memory parameter devised by Veitch and Abry (1999). We compare this estimator with two semiparametric estimators in the spectral domain, the local Whittle (LW) estimator developed by Robinson (1995a) and the “log– periodogram” (LP) estimator by Geweke and Porter–Hudak (1983). The wavelet estimator performs well for a wide range of nonlinear long–memory processes in the conditional mean and the conditional variance, and is reliable for discriminating between change–points and long–range dependence in volatility. We also address the issue of selection of the range of octaves used as regressors by the weighted least squares estimator. We will see that using the feasible optimal bandwidths for either the LW and LP estimators is a useful rule of thumb for selecting the lowest octave. We apply the wavelet estimator to volatility series of high frequency (intra–day) FX rates, and to the volatility and volume of stocks of the Dow Jones Industrial Average Index.
We propose and study by means of simulations and graphical
tools a class of goodness-of-fit tests
for ARCH models. The tests are based on the empirical
distribution function of squared residuals and smooth (parametric)
bootstrap. We examine empirical size and power by means of a
simulation study. While the tests have overall correct size, their
power strongly depends on the type of alternative and is particularly
high when the assumption of Gaussian innovations is violated.
As an example, the tests are applied to returns on Foreign Exchange
rates.
Key words: ARCH model, empirical process,
goodness-of-fit tests, size-power curves, smooth bootstrap,
squared residuals.
The paper deals with the power and robustness of the \(R/S\)-type tests under “contiguous” alternatives. We briefly review some long memory models in levels and volatility, and describe the \(R/S\)-type tests used to test for the presence of long memory. The empirical power of the tests is investigated when replacing the fractional difference operator \( (1−L)^d \) by the operator \( (1−rL)^d\), with \( r < 1 \) close to 1, in the FARIMA, LARCH and ARCH time series models. We also investigate the Gegenbauer process with a pole of the spectral density at frequency close to zero.
We consider a class of microeconomic models with interacting agents
which replicate the main properties of asset prices time series:
non-linearities in levels and common degree of long-memory in the
volatilities and co-volatilities of multivariate time series. For these
models, long-range dependence in asset price volatility is the
consequence of swings in opinions and herding behavior of market
participants, which generate switches in the heteroskedastic structure
of asset prices. Thus, the observed long-memory in asset prices
volatility might be the outcome of a change-point in the conditional
variance process, a conclusion supported by a wavelet analysis of the
volatility series. This explains why volatility processes share only
the properties of the second moments of long-memory processes, but not
the properties of the first moments.
Key words: long-memory, field effects,
interaction models, change-points, wavelets.
We derive the asymptotic distribution of the sequential empirical process of the squared residuals of an ARCH(\(p\)) sequence. Unlike the residuals of an ARMA process, these residuals do not behave in this context like asymptotically independent random variables, and the asymptotic distribution involves a term depending on the parameters of the model. We show that in certain applications, including the detection of changes in the distribution of the unobservable innovations, our result leads to asymptotically distribution free statistics.
The paper is concerned with the estimation of the long memory parameter in a
conditionally heteroskedastic model proposed by Giraitis et al. (1999b). We consider
estimation methods based on the partial sums of the squared observations, which are
similar in spirit to the classical \(R / S\) analysis, as well as spectral domain
approximate maximum likelihood estimators. We review relevant theoretical results
and present an empirical simulation study.
Key words: long memory, ARCH models, semiparametric
estimation, modified \(R/S\), KPSS and \(V/S\) statistics, periodogram.
We estimate here several trivariate FIGARCH models on three series of intra-day FX
rates returns: USD/DEM, USD/GBP, and USD/JPY.
We consider the trivariate constant conditional correlation CCC-FIGARCH, the
unrestricted trivariate FIGARCH, and a trivariate double long-memory model
combining an ARFIMA regression function with an unrestricted trivariate FIGARCH
skedastic function. Estimation results show that: (i) the three series are
anti-persistent and share a common degree of short-range dependence,
(ii) the series USD/DEM and USD/JPY have the same regression function,
(iii) the three series share the same degree of long-memory in their conditional
variance,
(iv) the conditional covariances \(\mbox{Cov}_t(\mbox{USD/DEM, USD/JPY})\)
and \(\mbox{Cov}_t(\mbox{USD/DEM,USD/GBP})\) have a common degree of persistence,
although this degree is different from the degree of long-memory of the conditional
variances,
(v) the unrestricted FIGARCH model dominates the CCC-FIGARCH model, although
(vi) the seasonality in the volatility of these series cannot be captured by FIGARCH
models.
Key words: intra-day data, long-memory,
antipersistence, heteroskedasticity, multivariate models, multivariate unconstrained
FIGARCH model, double long-memory, multivariate ARFIMA-FIGARCH models, second
generation models.
In 1996, I pioneered the class of double long memory processes:
Double Long-Memory Financial Time Series (1996) Univ. of London, with the ARFIMA-FIGARCH. This was my first paper in time series analysis.
Presented in the 1997 Econometric Society European Meeting, the 1997 Society for Economic
Dynamics conference Oxford, the 28th Workshop of the Euro Working Group on Financial Modelling, May 2001,
Slides
Long-Memory in Economics
G. Teyssière and A. Kirman editors, Springer (2007).
MR2263582
Zbl 1106.91001
Dependence in Probability and Statistics,
Lecture Notes in Statistics, Vol 200.
G. Lang, D. Surgailis and G. Teyssière editors, Springer (2010).
MR2741808
Zbl 1196.60011