Andreia Hall

Inferential Performance Assessment of Stochastic Optimisers and the Attainment Function

The performance of stochastic optimisers can be assessed experimentally on given problems by performing multiple optimisation runs, and analysing the results. Since an optimiser may be viewed as an estimator for the (Pareto) minimum of a (vector) function, stochastic optimiser performance is discussed in the light of the criteria applicable to more usual statistical estimators. Multiobjective optimisers are shown to deviate considerably from standard point estimators, and to require special statistical methodology. The attainment function is formulated, and related results from random closed-set theory are presented, which cast the attainment function as a mean-like measure for the outcomes of multiobjective optimisers. Finally, a covariance-measure is defined, which should bring additional insight into the stochastic behaviour of multiobjective optimisers. Computational issues and directions for further work are discussed at the end of the paper.

Standardization of a phonetic-phonological test for European-Portuguese children.

Objective: To develop and standardize a phonetic-phonological test (Teste Fonético-Fonológico-Avaliação da Linguagem Pré-Escolar, TFF-ALPE) for the assessment of European-Portuguese (EP) children’s articulation and phonological abilities. Patients and Methods: In order to standardize TFF-ALPE, 768 children aged 3;0–6;11 participated in this study. The standardization, validity and reliability of TFF-ALPE were analyzed. Results: TFF-ALPE presents strong cohesion and has strong inter- and intrajudge reliability. There was also a strong correlation between the TFF-ALPE data and those obtained in other studies. The content validity was demonstrated by the description of the test domain and the items that comprise TFF-ALPE. Conclusion: TFF-ALPE is a valid and reliable phonetic-phonological assessment instrument that speech-language pathologists can use with EP-speaking children.

Intelligibility as a Clinical Outcome Measure Following Intervention with Children with Phonologically Based Speech-Sound Disorders.

BACKGROUND The effectiveness of two treatment approaches (phonological therapy and articulation therapy) for treatment of 14 children, aged 4;0-6;7 years, with phonologically based speech-sound disorder (SSD) has been previously analysed with severity outcome measures (percentage of consonants correct score, percentage occurrence of phonological processes and phonetic inventory). Considering that the ultimate goal of intervention for children with phonologically based SSD is to improve intelligibility, it is curious that intervention studies focusing on childrens phonology do not routinely use intelligibility as an outcome measure. It is therefore important that the impact of interventions on speech intelligibility is explored. AIMS This paper investigates the effectiveness of the two treatment approaches (phonological therapy and articulation therapy) using intelligibility measures, both in single words and in continuous speech, as the primary outcome. METHODS & PROCEDURES Fourteen children with phonologically based SSD participated in the intervention. The children were randomly assigned to phonological therapy or articulation therapy (seven children in each group). Two assessment methods were used for measuring intelligibility: a word identification task (for single words) and a rating scale (for continuous speech). Twenty-one unfamiliar adults listened and judged the childrens intelligibility. Reliability analyses showed overall high agreement between listeners across both methods. OUTCOMES & RESULTS Significant improvements were noted in intelligibility in both single words (paired t(6)=4.409, p=0.005) and continuous speech (asymptotic Z=2.371, p=0.018) for the group receiving phonology therapy pre- to post-treatment, but no differences in intelligibility were found for those receiving the articulation therapy pre- to post-treatment, either for single words (paired t(6)=1.763, p=0.128) or continuous speech (asymptotic Z=1.442, p=0.149). CONCLUSIONS & IMPLICATIONS Intelligibility measures were sensitive enough to show changes in the phonological therapy group but not in the articulation therapy group. These findings emphasize the importance of using intelligibility as an outcome measure to complement the results obtained with other severity measures when exploring the effectiveness of speech interventions. This study presents new evidence for the effectiveness of phonological therapy in improving intelligibility with children with SSD.

Extremes of integer-valued moving average models with regularly varying tails

In this paper we study the limiting distribution of the maximum term of non-negative integer-valued moving average sequences of the form Xn = ∑i = −∞∞ βi ○ Zn − i where { Zn} is an iid sequence of non-negative integer-valued random variables with regularly varying tails, βi○Zn−i denotes binomial thinning. Several models are considered allowing different dependence structures on the thinning operations. For these models we manage to establish results which are similar to the ones obtained for the classic linear moving average: { Xn} > has an extremal index θ = \maxi{ βiα} / (∑βiα), where α is the index of regular variation, irrespectively of the dependence structure considered. The paper is concluded with a simulation study to illustrate the results.

Temporal acoustic correlates of the voicing contrast in European Portuguese stops

This study focuses on the temporal analysis of stops /p b t d k g/ and devoicing analysis of voiced stops /b d g/ produced in different word positions by six native speakers of European Portuguese. The study explores acoustic properties related to voicing. The following acoustic properties were measured: voice onset time (VOT), stop duration, closure duration, release duration, voicing into closure duration, duration of the preceding vowel and duration of the following vowel. Results suggested that when [b d g] were devoiced, the acoustic properties stop duration, closure duration, duration of the following vowel, duration of the preceding vowel and duration of voicing into closure were relevant for the voicing distinction. Implications for research and practice in speech and language therapy are discussed. Further investigation is needed to find how the productions analysed in the present study were perceived by listeners, specifically productions of devoiced stops.

Extremes of sub-sampled integer-valued moving average models with heavy-tailed innovations

Let {Xk} be a non-negative integer-valued stationary moving average sequence and define Yk=XTk as the sub-sampled series at a fixed integer interval T>1. We look at the limiting distribution of sample maxima of {Yk} and the corresponding extremal index.

Are crackles an appropriate outcome measure for airway clearance therapy

BACKGROUND: There is an urgent need to develop new outcome measures for respiratory therapy, to evaluate its effectiveness. Adventitious sounds generated from the lungs (crackles and wheezes), can now be quantified and characterized objectively with computer technology. To our knowledge, this is the first reported study designed to assess any change in lung crackles before and after a single session of airway clearance therapy. METHODS: Twenty-three stable bronchiectasis patients were recruited from United Kingdom out-patient clinics and treated with a single session of airway clearance therapy, using the active cycle of breathing technique. Sound recordings were made before and after the session at 7 anatomical chest locations. Computerized lung sound analysis was used to measure crackle parameters: 2-cycle deflection width (2CD), and crackle number per breath cycle (nBC). Perceived breathlessness, lung function, and oxygen saturation data were also recorded. RESULTS: Crackle mean 2CD and mean nBC increased post intervention. Sixteen participants (70%) showed a statistically significant difference in mean crackle 2CD before and after the session at ≥ 1 chest location. Thirteen (57%) participants had a difference between mean crackle 2CD before and after the intervention > 1 Smallest Real Difference (SRD, mean SRD = 2.23 ms) at ≥ 1 chest location. Differences in mean crackle nBC before and after the intervention did not exceed the SRD (mean SRD = 32 crackles per breath cycle) in any participant. Perceived breathlessness was significantly reduced post intervention; no significant changes were observed in either lung function or oxygen saturation. CONCLUSIONS: Crackle duration (2CD) was found to change after a single session of airway clearance therapy, and shows promise as a new outcome measure for respiratory therapy interventions.

Extremes of Stationary Sequences with Failures

Let {X n } be a stationary sequence subject to failures, meaning that some of the variables will be either withdrawn or somehow replaced. We discuss some extremal properties of three different failure models assuming that initially {X n } has an extremal index and satisfies appropriate dependence conditions. Special care is taken whenever X n is integer valued.

On the Maximum Term of MA and Max-AR Models with Margins in Anderson's Class

We consider several integer-valued stationary models of MA and max-AR type and study the limiting distribution of the maximum term after appropriate normalization. In particular, we consider marginal distributions which do not belong to the domain of attraction of any extreme value distribution but exhibit a quasi-stable limiting behavior in the sense of Anderson [J. Appl. Probab., 7 (1970), pp. 99–113] and therefore belong to the domain of attraction of a max-semistable law. Examples of such distributions are the negative binomial and the logarithmic distribution which are widely used to model real data applications. By verifying appropriate dependence conditions we obtain the limiting distribution of the maximum term for the models under consideration. Motivation comes from the analysis of the extremal behavior of integer-valued data requiring specific time series modeling.

Is the Portuguese version of the passage ‘The North Wind and the Sun’ phonetically balanced?

There is no standard phonetically balanced short passage for Portuguese research and clinical practice. This paper presents results of a novel analysis of ‘The North Wind and the Sun’ (NWS) passage that aims to determine if it is phonetically balanced for European Portuguese (EP) and Brazilian Portuguese (BP), based on new transcriptions resulting from algorithms developed for grapheme–phone transcription in these two varieties of Portuguese. These algorithms (based on standard EP and Sao Paulo BP varieties) are the same as those used to collect the frequency data, which is central to determining if a text is phonetically balanced. Results showed that neither transcription violates phonotactic rules, i.e. permissible combinations of speech sounds. The NWS is not phonetically balanced for BP if the phonemes are considered individually but is evenly distributed in terms of manner of articulation. The EP version of the NWS passage is a phonetically balanced text for EP.