PROFILE
Anani. Lotsi , BA (Legon,Ghana), MSc (Kaiserslautern, Germany), PhD(Groningen, The Netherlands))
Holds a PhD degree in Statistics from University of Groningen, The Netherlands and an Msc degree from University of Kaiserslautern from Germany. In addition he holds a BA degree in Computer Science and Statistics from university of Ghana, legon. He has over ten (10) years’ experience in teaching and research. Courses taught include Probability theory, Sample survey, Data analysis, Research Methodology, Actuarial Mathematics to mention but few. His research interest is broadly on
(i) network inference, (ii) high-dimensional inference, design and model choice (iii) Mixture model and (iv) Graphical modelling, Bid data analysis. His work has led to a number of scientific publications in international, peer-reviewed statistical and probability journals. He has also presented his work at several international conferences.
RESEARCH INTEREST/RESEARCH AREA
My research interests focus on (i) network inference, (ii) high-dimensional inference (iii) design and model choice (iv) graphical modelling (v) mixture modelling with applications in biology,(iv) deep learning and in scalable approaches to big data , (vii) big data management.
Other research interest include Variable selection, high-dimensional statistical learning, graphical models, model selection, theory and practice
PROFESSIONAL EXPERINECE
March 2015-Present: Department of Statistics, University of Ghana, Legon. Lecturer (Full Time)
May 2015-Present: Examiner for Institute for Chartered Accountancy Ghana (ICAG).
ACADEMICS
2010-2014: PhD in Statistics. University of Groningen. Thesis tittle: State space and graphical models for estimating
netwroks dynamics.
2002-2004: Msc.(Thesis option) in Mathematical Finance, University of Kaisesrlautern, Germany.
1998-2002: BA (Hons) in Computer Science and Statistics, University of Ghana Legon.
SELECTED PUBLICATIONS
(1 ) Lotsi, A. and Wit, E. (2016). State-space modeling of dynamic genetic networks. British Journal of Applied Science & Technology,17(4), pp.1-18
Abstract: The genomic reality is highly complex and dynamic. Recent developments of high-throughput technologies have enabled researchers to measure the RNA abundance of thousands of genes simultaneously. The challenge is to unravel from such measurements genomic interactions and key biological features of cellular systems. Two common problems are the high-dimensionality of the system and the spurious correlations induced by unmeasured intermediate substrates. Furthermore most currently available models cannot deal with biological replicates. Our goal is to devise a method for inferring large transcriptional or gene regulatory networks from high-throughput data sources such as gene expression microarrays with potentially hidden states, such as unmeasured transcription factors (TFs).
Keywords: genomic interactions; microarray experiments; dynamic networks; state space representation; EM algorithm.
(2) Nanga, S. and Lotsi, A.(2016). A Comparison between Logistic Regression and K Nearest Neighbor in Modeling Mortality Amongst Children Under five Years in Ghana. Dama International Journal of Researchers 1(6), pp. 60-67.
Abstract: Child mortality is regarded as one of the most revealing measures of society’s ability to meet the needs of its people. The Millennium Development Goal 4 (MDG 4) advocates a reduction of under-five mortality rate by two-thirds between 1990 and 2015. The main objective of this study was to develop a validated set of statistical models and select the most appropriate model between logistic regression and K Nearest Neighbor to predict mortality among children under five and to compare the influence of selected risk factors on the probability of death before the age of 5 years among children in Ghana. The study revealed that the K Nearest Neighbor model was the most efficient in modeling Mortality in Children under five with a CCR of 83%. The Logistic Regression model will also do a good job at predicting mortality in children under five with a CCR of 81%. The highest educational level of mother, Age of mother at birth, Type of toilet facility used by family, alcohol consumption and the wealth index of family were discovered as the most important variables in predicting mortality amongst children under five in Ghana across both models.
Keywords: Logistic Regression, Neural Networks, Children under five years,
(3) Darkwah K.A, Nortey E.N.N and Lotsi, A.(2016). Estimation of the Gini coefficient for the lognormal distribution of income using the Lorenz curve. SpringerPlus 5:1196, DOI 10.1186/s40064-016-2868z.
Abstract: The main objective of the study is to compare the Newton–Cotes methods such as the
Trapezium rule, Simpson 1/3 rule and Simpson 3/8 rule to estimate the area under the Lorenz curve and Gini coefficient of income using polynomial function with degree 5. Comparing the Gini coefficients of income computed from the Polynomial function with degree 5 for the Trapezium, Simpson 1/3 and Simpson 3/8 methods using the relative errors showed that the trapezium rule, Simpson’s 1/3 rule and Simpson’s 3/8 rule show negative biases with the Simpson 1/3 rule yielding the lowest absolute relative true error of 4.230711 %.
Keywords: Lorenz curve, Gini coefficient, Newton–Cotes methods, Lognormal distribution, Income distribution
(4) Lotsi, A. and Wit, E. (2014).Network estimation in State Space Model with L1 regularization constraint.Submitted to “Computational Statistics and Data Analysis” Awaiting
Abstract: Networks have arisen as an attractive paradigm of genomic science ever since the introduction
of large scale genomic technologies which carry the promise of elucidating the relationship in functional genomics. Microarray technologies and related methods coupled with appropriate mathematical and statistical models have made it possible to identify dynamic regulatory networks by measuring time course expression levels of many genes simultaneously. However one of the challenges is the high-dimensional nature of such data coupled with the fact that these gene expression data are known not
to include various biological process. As genomic interactions are highly structured we aim to derive a method for inferring a sparse dynamic network in a high dimensional data setting. We assume that the observations are noisy measurements of gene expression in the form of mRNAs, whose dynamics can be described by some partially observed process. We build an input-dependent linear state space model including these hidden states and demonstrate how an incorporated L1 regularization constraint in an
Expectation-Maximization (EM) algorithm can be used to reverse engineer transcriptional networks from gene expression profiling data. This corresponds to estimating the model interaction parameters. In silico validation experiment is performed to evaluate the performance of our method and the proposed method is illustrated on time-course microarray data obtained from a well established t-cell data.
Keywords: genomic, gene expression, microarray, sparse , EM algorithm, state space modelresponse from reviewers)
(5) Lotsi A, and Wit E.(2016).High dimensional Sparse Gaussian Graphical Mixture Model. Afrika Statistika 11(2), pp. 1041–1059.
Abstract: This paper considers the problem of networks reconstruction from heterogeneous data using a Gaussian Graphical Mixture Model (GGMM). It is well known that parameter estimation in this context is challenging due to large numbers of variables coupled with the degenerate nature of the likelihood. We propose as a solution a penalized maximum likelihood technique by imposing an l1 penalty on the precision matrix. Our approach shrinks the parameters thereby resulting in better identi_ability and variable selection. We use the Expectation Maximization (EM) algorithm which involves the graphical LASSO to estimate the mixing coe_cients and the precision matrices. We show that under certain regularity conditions the Penalized Maximum Likelihood (PML) estimates are consistent. We demonstrate the performance of the PML estimator through simulations and we show the utility of our method for high dimensional data analysis in a genomic application.
Keywords: Graphical, Mixture, Lasso, Expectation Maximization.
(6) Lotsi, A. and Wit, E. (2014). R-package for High dimensional Sparse Gaussian Graphical Mixture Model. http://cran.r-project.org/web/packages/glassomix/index.html.
MPhil/MSc Research Topics 2014-2015
(1) Modelling the risk and causes of maternal mortality in GHANA using hierarchical models (2008-2013) by Kasim (Carnegie)
(2) Stochastic time series modelling for total number of defective parts of printed circuit boards in the manufacturing industry in Ghana (MPHIL
(3) Prediction of Poverty Incidence in Statistically underdeveloped countries by ERNEST ENYAN (MPHIL)
(4) Application of Numerical integration to stochastic estimation of the GINI coefficient by Kwasi A. Darkwah (Carnegie)
(5) Modelling Mortality Amongst Children Under Five years in Ghana: Comparism of different Modelling Techniques by SALIFU NANG (Mphil)
(6) Actuarial Technique for Pricing Mutual Health Care Insurance (MSC.)
(7) Time series analysis on sickle cell disease (SCD) in Koforidua, Ghana: By Steh Ayretey.
MPhil/MSc Research Topics 2016-2017
(8) Amen Agbenorhevi (10215306) Analyzing the effect of Macroeconomic Variables on Inflation in Ghana Using Distributed Lag Models (Mphil Actuarial Science)
(9) Dennis Lartey Quayesam (10505288) Stochastic Modelling of Stock Prices on the Ghana Stock Exchange (Mphil Actuarial Science)
(10) Israel Elorm Dzokoto (10063284). Comparative Approaches to handling missing data for both cross-sectional and longitudinal models. (Mphil Statistics)
(11) Amewu Damian Mawuli (10508813) Evaluation of academics programs in Statistics in Ghanaian Universities using PCA and Rough set and TOPSIS . (Mphil Statistics)
(12) John Kennedy Banini (10551688). Actuarial modelling of insurance claims severity and frequency in Ghana with mixed Poisson distribution; a case of Statte Insurance Company Limited.
(13) Mawutor Fleku (10071460) Generalized Bernouilli Model for correlated Binary responses: Application to the National Income dynamics study (NIDS) Datasets
(14) KORENTENG NANA STERLING (10551851) Collective risk modelling in national health insurance
(15) MUNKAILLA SULLE-MAN. (10303113) Modelling risk using a stochastic (LEE-CATER) and it effects on the defined benefit scheme (SSNIT)
(16) Nicholas Nkum (10551810) Portfolio Optimization using minimum variance line approach. (A case study of SSNIT)
(17) Seidu Abdulai. (10550599) Statistical Modelling of factors affecting health insurance claims in GHANA
CONFERENCES, WORKSHOPS AND SEMINARS
Jan 2013: Talk. Sparse Gaussian Graphical Mixture Model. YES Workshop, 28-29 January 2013, Eindhoven, Netherlands
Nov 2013: Poster. Network estimation in State Space Model with L1 regularization constraint.
Workshop, 31 Oct-1 Nov Groningen, Netherlands.
Nov 2012: Poster. Sparse Copula Gaussian Model for count data variables with covariates
Programme 40th Stochastics Meeting Lunteren, 14-16 November 2012, Lunteren, Netherlands.
June 2012: Poster. Modelling Sparse ODEs using penalized graphical models Parameter Estimation for Dynamical Systems, June 4 - 6, 2012. Einghoven,
Netherlands
Workshop on complex networks, Eindhoven, The Netherlands, 29/01/13- 01/02/13
Young statisticians meeting on C. H. D.S, Eindhoven, The Netherlands, 28/01/13 - 29/01/13
Teaching Assistant Experience
2012 Statistics (C) Johann Bernoulli Institute, University of Groningen, The Netherlands.
2011 Statistics (C) Johann Bernoulli Institute, University of Groningen, The Netherlands.
Skills
Languages