Explore chapters and articles related to this topic
Computational Methods for Bayesian Analysis
Published in Gary L. Rosner, Purushottam W. Laud, Wesley O. Johnson, Bayesian Thinking in Biostatistics, 2021
Gary L. Rosner, Purushottam W. Laud, Wesley O. Johnson
Many convergence diagnostics are available in the R packages coda and boa. Several of the routines that are part of coda are built into OpenBUGS. Aside from providing built-in tools for assessing convergence, both of these packages also include functions to simplify the process of loading saved iterates from an MCMC program into R. The purpose of moving MCMC iterates into an analysis environment like R is to be able to manipulate the output when creating summary Bayesian inferences and/or to create pretty pictures of numerical approximations to posterior and predictive distributions.
Towards a learning stance in teams: Developing a community of practice to capture and disseminate what works for whom
Published in Ilana B. Crome, Richard Williams, Roger Bloor, Xenofon Sgouros, Substance Misuse and Young People, 2019
The first section charts the relationship between evidence-based practice and the realities of working with youth who use substances. The next section describes the rationale for producing and using treatment manuals and presenting these documents as flawed necessities. Then a brief coda to describe mentalisation is introduced. Mentalisation is the core theoretical construct underpinning AMBIT, a voluntary sector project based at the Anna Freud National Centre for Children and Families, a charity based in London, that in the past two years has trained over 200 teams in AMBIT across the UK and internationally. The penultimate section describes, in more detail, the shortcomings of manuals, some of which AMBIT’s innovative approach to web-based, user-adaptable, treatment manuals is intended to address. The AMBIT approach to manualisation and learning in teams is described in the final section, which includes an overview of the web-based technology that supports AMBIT’s intention to connect workers and teams across geographical space and to support sharing emerging best practice.
Combinatorial and Model-Based Methods in Structuring and Optimizing Cluster Trials
Published in Zoran Antonijevic, Robert A. Beckman, Platform Trial Designs in Drug Development, 2018
Valerii V. Fedorov, Sergei L. Leonov
In more complex cases, clusters may be defined by a combination of several categories (genotypes, phenotypes, mutational signatures, etc.) that can be determined with various statistical methods like cluster analysis, principal component and factor analysis, or other machine learning or statistical techniques which are beyond the scope of this chapter; cf. De Souto et al. (2008), Alexandrov et al. (2013), Prat et al. (2013), Iorio et al. (2016). It is worth noting that the term “cluster” may be overloaded: recall “cluster randomization,” as in Donner and Klar (2000), or “cluster cancer,” as in Thun and Sinks (2004). “Meta trial” or “compound trial” are competitive alternatives, but they are overloaded even more, so that we decided to stay with “cluster trials” for the needs of this chapter. We stay within the framework of oncology trials but the approach can be applied in other areas, for instance, for the comparison of different antibiotics for the same disease; cf. CODA, https://clinicaltrials.gov/ct2/show/NCT02800785.
Partial least squares regression with compositional response variables and covariates
Published in Journal of Applied Statistics, 2021
Jiajia Chen, Xiaoqin Zhang, Karel Hron
For the statistical analysis of CoDa, if we directly use the existing methods and ignore the compositional character of CoDa, it might lead, as is well known, to spurious results. In fact, CoDa are characterized by specific geometrical properties, represented by the Aitchison geometry [25]. The principle of working on coordinates is widely used in CoDa analysis, that is, express CoDa in log-ratio coordinates (coefficients) in real space [12,22]. Popular log-ratio coordinates (coefficients) are additive log-ratio (alr) coordinates [1], centered log-ratio (clr) coefficients [1] and isometric log-ratio (ilr) coordinates [11]. To preserve the invariance of distance from the simplex to real space, the clr coefficients and ilr coordinates are recommended. For any composition 1] are 5,24] have been proposed to deal with zeros.
Sound discrimination and explicit mapping of sounds to meanings in preschoolers with and without developmental language disorder
Published in International Journal of Speech-Language Pathology, 2021
Carolyn Quam, Holly Cardinal, Celeste Gallegos, Todd Bodner
It is possible that children ignored duration distinctions in the present study because they are not used contrastively in English (Dietrich, Swingley, & Werker, 2007). However, this is also true of pitch (Quam & Swingley, 2010). Duration is also critical to speech perception in English. It is a secondary cue to the voicing of coda consonants, and six-year-old (Krause, 1982) and adult listeners (Peterson & Lehiste, 1960) can exploit duration differences in consonant perception. The distance between the centres of the duration categories in our task was 450 ms. Prior work (Krause, 1982, Figure 1) indicates that 3-year-olds can distinguish duration differences of roughly 200 ms., and the average difference between vowel lengths preceding voiced vs. voiceless consonants in English is only 100 ms. (Peterson & Lehiste, 1960). Thus, future work using more naturalistic vowel sounds embedded in words and carrier phrases may find improved duration discrimination and mapping performance.7
Performance of the hotelling T2 control chart for compositional data in the presence of measurement errors
Published in Journal of Applied Statistics, 2019
F. S. Zaidi, P. Castagliola, K. P. Tran, M. B. C. Khoo
In the case of continuous multivariate processes, the vast majority of control charts assume that the data are unconstrained. But there is a specific category of multivariate data which are constrained by definition. This kind of data is called CoDa (for Compositional Data) and it is represented by vectors whose strictly positive components only convey relative information. CoDa includes measurements in probability, proportions, percentages and parts. Usually the sum of the components of CoDa vector is expressed as some constant κ being equal to 1 if we are working with proportions, 100 if we are working with percentages, 20,21] or [23]), in the food industry (see for instance [14]) and in the field of biology (see for instance [3]). Today, it is considered that a rigorous mathematical foundation of CoDa is due to Aitchison [1], who developed an adequate geometry to model and transform such data. This will be detailed in Section 2.