چند رسانه‌ای


 

bullet سخنرانی‌های گروه آمار

مقایسه شبیه سازی مونت کارلو و بوت استرپ

دکتر نصراله ایران پناه

دریافت فایل سخنرانی

 

پیشرفت های جدید در تعیین پیش های عینی

دکتر محمدرضا مشکانی

دریافت فایل سخنرانی

 

A quick overview of robust statistics, fundamental concepts and definitions

دکتر اشکان شباک پژوهشکده آمار ایران

دریافت فایل سخنرانی

Abstract
One of the first steps towards obtaining a coherent statistical analysis is the detection of correct outliers that may otherwise adversely lead to model misspecification, biased parameter estimation and incorrect results. Nevertheless, these atypical data may carry important information. Moreover, often happens in practice that although the majority of collected observations hold the assumptions of the model, some observations follow different pattern or no pattern at all. Then classical methods and analysis, which rely heavily on the assumptions that are being violated by outliers, often have very poor performance. The Robust statistical approach, seeks to provide an alternative approach to classical statistical methods that emulate popular statistical methods, but which are not unduly affected by outliers or other small departures from model assumptions. So in this lecturer we are going to introduce the fundamental concepts and some basic definitions of this area that rapidly becomes more and more important in statistics in recent decades, while a brief history of the robust statistics are given. Some applications of robust estimators in the statistical inference also will be discussed to the extent of 1 hour time permits.
Audiences: This scientific lecture may appropriate for statistics students, statisticians, researchers and anyone who is interested in statistical subjects.
Outlines:

  • Definitions of outliers, their effect and detection methods
  • Why do we need robust Statistics?
  • History of robust Statistics
  • Fundamental Concepts of the Robustness
  • Affine Equivariant Estimators of Location and Scale
  • Robust estimators
  • Application of robust statistics in regression and multivariate analysis

 

Penalized Likelihood Ratio Test for Generalized Linear Models with an Unknown biomarker Cut-point in Clinical Trials

دکتر پریسا گوانجی از دانشگاه کویینز کانادا

دریافت فایل سخنرانی

In clinical trials, new treatments may tend to benefit a subset of patients more. Biomarkers are frequently used to identify this subset of patients in order to avoid unnecessary therapies and failure to recognize beneficial treatments. We are interested in testing the biomarker effects in a generalized linear model to see if the new treatment benefits all patients in the same way or not. When the subset is defined by a continuous biomarker with an unknown cut–point, the regularity conditions of the ordinary likelihood ratio test are not satisfied. We first approximate the indicator function, defining biomarker subgroups, by a smooth continuous function. To overcome irregularities, we develop a penalized likelihood method, introducing a new idea of using random penalty term. Proposing a new set of regularity conditions helps us to study the properties and limiting distributions of the maximum penalized likelihood estimates of the parameters. We further prove that the penalized likelihood ratio test statistic has an asymptotic chi–square distribution with degrees of freedom under the null hypothesis. Extensive simulation studies show that the proposed test procedure works well for hypothesis testing. The proposed method is applied to a clinical trial of DIG data, considering the effect of Digoxin on mortality and morbidity in patients with heart failure, to show that Digoxin is more effective (in terms of reduction in death or hospitalization from worsening heart failure) on patients with higher ejection fraction level.

برآورد ارتباطات کارکردی بین نواحی مغزی با استفاده از مدل های نوسانات چندمتغیره

دکتر رحمان فرنوش از دانشگاه علم و صنعت

دریافت فایل سخنرانی

برآورد ارتباط های بین نواحی مغزی و کشف شبکه های مغزی از جمله موضوعات مورد توجه محققان در این دهه است. روش های آماری زیادی از جمله سری های زمانی نقش مهمی در برآورد این ارتباط ها دارند. در این راستا مدل های نوسانات چندمتغیره از جمله مدل میانگین متحرک بصورت نمایی وزن دار شده و مدل همبستگی شرطی پویا، برای برآورد ارتباط های کارکردی پویا استفاده می شوند. با توجه به آنکه ارتباط های واقعی بین نواحی مغزی کشف نشده است، جهت ارزیابی عملکرد این مدل ها از شبیه سازی استفاده می گردد. در این تحقیق، ابتدا نقاط ضعف مطالعات شبیه سازی پیشین بیان و اصلاح می گردد. سپس با استفاده از مانده های استاندارد شده، توانایی مدل های فوق در برآورد ارتباط های کارکردی افزایش می یابد. در آخر مدلی که کمترین میزان میانگین مربع خطا را دارد به عنوان بهترین مدل انتخاب شده و برای برآورد ارتباط های کارکردی نواحی مغزی در داده های تصویربرداری در حالت استراحت ایرانی مورد استفاده قرار می گیرد.

مدل بندی جمعی توابع چگالی و کاربرد آن در مسائل کلاس‌بندی و پیشگویی ساختار پروتئین

توسط جناب آقای دکتر مهدی معدولیت از دانشگاه مارکوئیت ایالات متحده در تاریخ 11 خرداد ماه 1395 برگزار شد.
دریافت فایل سخنرانی

Abstract: "This talk develops a method for simultaneous estimation of density functions for a collection of populations of protein backbone angle pairs using a data-driven, shared basis that is constructed by bivariate spline functions defined on a triangulation of the bivariate domain. The circular nature of angular data is taken into account by imposing appropriate smoothness constraints across boundaries of the triangles. Maximum penalized likelihood is used to fit the model and an alternating block-wise Newton-type algorithm is developed for computation. A simulation study shows that the collective estimation approach is statistically more efficient than estimating the densities individually. The proposed method was used to estimate neighbor-dependent distributions of protein backbone dihedral angles (i.e., Ramachandran distributions). The estimated distributions were applied to protein loop modeling, one of the most challenging open problems in protein structure prediction, by feeding them into an angular-sampling-based loop structure prediction framework. Our estimated distributions compared favorably to the Ramachandran distributions estimated by fitting a hierarchical Dirichlet process model; and in particular, our distributions showed significant improvements on the hard cases where existing methods do not work well."
Keywords: "Bivariate splines, Log-spline density estimation, Protein structure, Ramachandran distribution, Roughness penalty

پلی ما بین فرآیندهای نقطه‌ای فضایی و علوم شناختی

توسط سرکار خانم دکتر فرزانه صفوی منش از دانشگاه شهید بهشتی در تاریخ 4 اسفند ماه 1394 برگزار شد.
دریافت فایل سخنرانی

Abstract:Analyzing point patterns with linear structures has recently been of interest in e.g. neuroscience and geography. To detect anisotropy in such cases, we introduce a functional summary statistic, called the cylindrical K-function, since it is a directional K-function whose structuring element is a cylinder. Further we introduce a class of anisotropic Cox point processes, called Poisson line cluster point processes. The points of such a process are random displacements of Poisson point processes defined on the lines of a Poisson line process. Parameter estimation based on moment methods or Bayesian inference for this model is discussed when the underlying Poisson line process and the cluster memberships are treated as hidden processes. To illustrate the methodologies, we analyze a two and a three-dimensional point pattern data set. The 3D data set is of particular interest as it relates to the minicolumn hypothesis in neuroscience, claiming that pyramidal and other brain cells have a columnar arrangement perpendicular to the pial surface of the brain.

پیش‌بینی موقعیت استقرار پروتئین‌ها در اجزای زیرسلولی با استفاده از رویکرد پیشنهاد شخصی براساس شبکه‌های دوبخشی

توسط جناب آقای دکتر چنگیز اصلاحچی از دانشگاه شهید بهشتی در تاریخ 28 فروردین ماه 1395 برگزار شد.
دریافت فایل سخنرانی

Motivation: The importance of protein subcellular localization problem is due to the importance of protein’s functions in different cell parts. Moreover Prediction of subcellular locations helps to identify the potential molecular targets for drugs and has an important role in genome annotation. Most of the existing prediction methods assign only one location for each protein. Since proteins move between different subcellular locations, each protein can have multiple subcellular locations. In recent years, some multiple location predictors have been introduced. However, their performances are not accurate enough and there is much room for improvement. Results: In this paper we introduced a method, pmLplr, to predict locations for a protein. In pmLplr, we use a personal recommender method, NBI, to tackle the location prediction problem. In order to predict locations for each protein, the similarity of the proteins considered. This similarity is derived from STRING, protein-protein interaction database. PmLplr predicts a list of locations for each protein and it can properly overcome the multiple location prediction problem. For evaluating the performance of PmLplr, we considered two datasets, proteins of HUMAN and RAT. The performance of this algorithm is compared with three state-of-the-art algorithms, YLoc, WOLF-PSORT and prediction channel. The results indicate that our proposed method is significantly superior compare to the mentioned methods.

Variable selection in finite mixture of survival models for biomedical genomic studies.

دکتر فرهاد شکوهی از دانشگاه مک‌گیل کانادا ، مورخ 03/07/95

دریافت فایل سخنرانی

 

تاریخچه نظریه اطلاع

دکتر غلامرضا محتشمی برزادران از دانشگاه فردوسی مشهد، مورخ 28/07/95

دریافت فایل سخنرانی

 

اندازه گیری گذر از مدرسه به کار

دکتر فرهاد مهران از دانشگاه نوشاتل سوئیس، مورخ 30/09/95

دریافت فایل سخنرانی

 

مدلسازی مطالعات طولی

"دکتر سمانه افتخاری مهابادی" عضو هیأت علمی دانشگاه تهران، مورخ 26/11/95

دریافت فایل سخنرانی

 

Incorporating Prior Information into the Problem of Multiple Hypothesis Testing with Application to the Analysis of Genetic Association Data

"دکترعلی کریم‌نژاد" عضو هیأت علمی دانشگاه صنعتی خواجه نصیرالدین طوسی، 95/12/17
دریافت فایل سخنرانی

An important objective in simultaneous hypothesis testing terminology is to control false discoveries. In most of studies in the literature, local false discovery rates (LFDRs) are estimated based on a combined analysis in which all features presented together are combined to be analyzed together. However, sometimes the nature of data or presence of some covariates suggests a separate analysis in which features are assigned to some reference classes, leading to more reliable estimates of LFDRs. Estimating LFDRs corresponding to single nucleotide polymorphisms (SNPs) associated with the coronary artery disease is a promising example in which, based on some available biological information, SNPs can be assigned to some biological reference classes such as exonic and ncRNA. Conducting both separate and combined analyses might lead to two different decisions based on estimated values of LFDRs. Thus, it is a debatable issue to determine if any of the SNPs is associated with the disease. In this talk, we introduce novel approaches including robust Bayes and information-theoretic approaches to estimate LFDRs, overcoming with the uncertainty in making a wise decision based on both separate and combined analysis.

Confidence Interval in Randomized Nomination Sampling

"دکتر محمد نورمحمدی" رئیس پژوهشکده آمار، مورخ 96/02/12

دریافت فایل سخنرانی

 

Machine Learner Fusion for Regression Problems

"دکتر علی شمس‌الدینی" عضو هیأت علمی دانشگاه تربیت مدرس، مورخ 96/03/02

دریافت فایل سخنرانی