Time filter

Source Type

Araujo A.,Competence Center for Methodology and Statistics | Meira-Machado L.,University of Minho | Roca-Pardinas J.,University of Vigo
Journal of Statistical Software

One major goal in clinical applications of multi-state models is the estimation of transition probabilities. The usual nonparametric estimator of the transition matrix for nonhomogeneous Markov processes is the Aalen-Johansen estimator (Aalen and Johansen 1978). However, two problems may arise from using this estimator: first, its standard error may be large in heavy censored scenarios; second, the estimator may be inconsistent if the process is non-Markovian. The development of the R package TPmsm has been motivated by several recent contributions that account for these estimation problems. Estimation and statistical inference for transition probabilities can be performed using TPmsm. The TPmsm package provides seven different approaches to three-state illnessdeath modeling. In two of these approaches the transition probabilities are estimated conditionally on current or past covariate measures. Two real data examples are included for illustration of software usage. © 2014, American Statistical Association. All rights reserved. Source

Renard L.M.,Center for Health Studies | Renard L.M.,University of Paris Descartes | Bocquet V.,Competence Center for Methodology and Statistics | Vidal-Trecan G.,University of Paris Descartes | And 3 more authors.

Introduction: Type 2 diabetes is associated with severe micro- and macro-vascular complications. Physicians' and patients' adherence to follow-up guidelines permits postponing or reducing these complications. The objectives were to assess the level of adherence to fundamental follow-up guidelines and determine patients' characteristics associated with this level of adherence in the context of Luxembourg, where no guidelines were implemented. Study population: The exhaustive residing population treated for type 2 diabetes in Luxembourg during the 2000-2006 period (N = 21,068). Methods: Seven fundamental criteria were extracted from international guidelines (consultation with the treating physician, HbA1c tests, electrocardiogram, retinal, dental, lipid and renal check-ups). The factors associated with the level of adherence to those criteria were identified using a partial proportional odds model. Results: In 2006, despite 90% of the patients consulted at least 4 times their treating physician, only 0.6% completed all criteria; 55.0% had no HbA1c test (28.6 points since 2000) and 31.1% had a renal check-up (+21.6 points). The sex (ORmale: 0.87 [95%CI, 0.8320.92]), the nationality (OR NonEU: 0.64 [0.5220.78]), the type of antidiabetic treatment (OR oral: 1.48 [1.3521.63], ORmixed: 1.35 [1.202 1.52]) and the type of treating physician (ORG-ID: 0.47 [0.4220.53]) were the main factors associated with the level of adherence in 2006 (3 or more criteria). Conclusion: A large percentage of patients were not provided with a systematic annual follow-up between 2000 and 2006. This study highlighted the necessity to promote guidelines in Luxembourg, education for physicians and to launch a national discussion on a disease management program for diabetic patients. © 2013 Renard et al. Source

Senn S.,Competence Center for Methodology and Statistics

I compare two possible extreme hypotheses regarding submission of papers to journals: the Q hypothesis, whereby the decision to submit is based on quality of research; and the P hypothesis, whereby it is based on probability of acceptance. I give five reasons as to why the P hypothesis is more plausible and suggest that problems of missing data may previously have caused researchers to misinterpret the evidence on editorial bias. © 2013 Senn S. Source

Hipp G.,Center Hospitalier Of Luxembourg | Diederich N.J.,Center Hospitalier Of Luxembourg | Diederich N.J.,University of Luxembourg | Pieria V.,Center Hospitalier Of Luxembourg | Vaillant M.,Competence Center for Methodology and Statistics
Journal of the Neurological Sciences

Background In early stages of idiopathic Parkinson's disease (IPD), lower order vision (LOV) deficits including reduced colour and contrast discrimination have been consistently reported. Data are less conclusive concerning higher order vision (HOV) deficits, especially for facial emotion recognition (FER). However, a link between both visual levels has been hypothesized. Objective To screen for both levels of visual impairment in early IPD. Methods We prospectively recruited 28 IPD patients with disease duration of 1.4 +/- 0.8 years and 25 healthy controls. LOV was evaluated by Farnsworth-Munsell 100 Hue Test, Vis-Tech and Pelli-Robson test. HOV was examined by the Ekman 60 Faces Test and part A of the Visual Object and Space recognition test. Results IPD patients performed worse than controls on almost all LOV tests. The most prominent difference was seen for contrast perception at the lowest spatial frequency (p = 0.0002). Concerning FER IPD patients showed reduced recognition of "sadness" (p = 0.01). "Fear" perception was correlated with perception of low contrast sensitivity in IPD patients within the lowest performance quartile. Controls showed a much stronger link between "fear" perception" and low contrast detection. Conclusion At the early IPD stage there are marked deficits of LOV performances, while HOV performances are still intact, with the exception of reduced recognition of "sadness". At this stage, IPD patients seem still to compensate the deficient input of low contrast sensitivity, known to be pivotal for appreciation of negative facial emotions and confirmed as such for healthy controls in this study. © 2014 The Authors. Source

Senn S.,Competence Center for Methodology and Statistics

In analysing whether there is an editorial bias in favour of positive studies, researchers have made implicit assumptions that are implausible. In particular, to justify the conclusion that there is no bias because observed editorial acceptance rates do not favour positive studies, the assumption that the decision to submit an article is based solely on quality would be required. If, on the other hand, submission were based on perceived probability of acceptance, negative and positive studies would not differ in terms of acceptance rates, but in terms of quality. It is shown, using a simple graphical model, how similar underlying situations as regards the relationship between quality and probability of acceptance on the one hand and study outcome (positive or negative) and probability of acceptance on the other could produce dramatically different results depending on the behaviour of authors. Furthermore, there is, in fact, some evidence that submitted negative studies are, on average, of higher quality than positive ones. This calls into question the standard interpretation of the studies examining editorial bias. It would appear that despite similar probabilities of acceptance for negative and positive studies, editors could be discriminating against negative studies. © 2012 Senn S. Source

Discover hidden collaborations