Steven D.

Statistics PhD with Years of Experience

Statistics PhD with Years of Experience

Actuarial Science

Actuarial Science is scientific field which focuses on modeling insurance risks and managing practical aspects of those. Actuaries are closely linked to financial engineers working on pricing and risk management of credit/mortgage derivatives as well as biostatisticians running survival analyses for patients and other subjects. The three fields employ similar methods and borrow fresh ideas from one another. Of course, there are implementational nuances in each field. There are different ways of monetizing the adverse risks in each field. Actuarial science can be split into two major directions: 1) getting the exact risk profile of a single client, based on his/her characteristics, 2) modeling correlated claims from multiple clients in the same insurance portfolio. I have more than ten years of academic and practical experience in each direction. From the mathematical point of view, the correlated nature of different claims (direction 2) poses the most interesting mathematical challenges. Still, the day-to-day juice filling the salary of an actuary is direction 1: understanding how the risk profile of client A is different from the risk profile of client B as a function of time. My past actuarial projects have included building models for correlated credit risks using standard and non-standard copulas, applying Kendall's tau and Spearman's rho to prescreen relationships between selected risk factors, building a survival curve for a specific segment of subjects using non-parametric smoothing techniques and model-based interpolation, applying extreme value theory to extrapolate survival curves into durations where the data are sparse, running cluster analysis to see if it confirms domain-knowledge based segmentation of subjects into different risk profiles, performing extensive Monte Carlo simulation of relative simple models and identifying situations where their errors are relatively small, testing ad-hoc adjustments implemented in real analytic systems before me, etc. The implementation has been done in MATL

Econometrics

Over the last eleven years I have worked with more than two hundred economists, social scientists, finance professionals and formally introduced finance companies, including hedge funds. We have dealt with problems pertaining to time series analysis, heteroskedastic models, panel data, structural equation modeling, robust inference, bootstrap, multi-factor representations, etc... Stata, R and EViews were my preferred tools for tackling econometric problems but I also gave clients the option to work in Matlab, SAS or Python. In my humble opinion, JMP, SPSS and Minitab are somewhat simplistic when it comes to serious econometric tasks. They are more suitable for other applied areas, like biology or quality control.

Actuarial Science

Actuarial Science is scientific field which focuses on modeling insurance risks and managing practical aspects of those. Actuaries are closely linked to financial engineers working on pricing and risk management of credit/mortgage derivatives as well as biostatisticians running survival analyses for patients and other subjects. The three fields employ similar methods and borrow fresh ideas from one another. Of course, there are implementational nuances in each field. There are different ways of monetizing the adverse risks in each field. Actuarial science can be split into two major directions: 1) getting the exact risk profile of a single client, based on his/her characteristics, 2) modeling correlated claims from multiple clients in the same insurance portfolio. I have more than ten years of academic and practical experience in each direction. From the mathematical point of view, the correlated nature of different claims (direction 2) poses the most interesting mathematical challenges. Still, the day-to-day juice filling the salary of an actuary is direction 1: understanding how the risk profile of client A is different from the risk profile of client B as a function of time. My past actuarial projects have included building models for correlated credit risks using standard and non-standard copulas, applying Kendall's tau and Spearman's rho to prescreen relationships between selected risk factors, building a survival curve for a specific segment of subjects using non-parametric smoothing techniques and model-based interpolation, applying extreme value theory to extrapolate survival curves into durations where the data are sparse, running cluster analysis to see if it confirms domain-knowledge based segmentation of subjects into different risk profiles, performing extensive Monte Carlo simulation of relative simple models and identifying situations where their errors are relatively small, testing ad-hoc adjustments implemented in real analytic systems before me, etc. The implementation has been done in MATL

Biostatistics

Over the last eleven years I have worked with more than a hundred of PhDs, MDs, pharmaceutical professionals and biologists on problems pertaining to spatial patterns of species in different remote areas, intra-year growth dynamics, GWAS, epidemiology and drug assessment. The posed problems called for a collection of statistical methods each time. The examples were generalized linear models, zero inflation, multivariate analysis of variance, survival analysis, latent state detection, propensity score matching, spatial statistics, ... Typically, genetics called for high-tech statistical weaponry while drug testing could be done using more traditional statistical tools. The following statistical packages were used in the process: R, Minitab, SAS, JMP, SPSS and Matlab. I have a PhD in Statistics (Berkeley).

Biostatistics

Over the last eleven years I have worked with more than a hundred of PhDs, MDs, pharmaceutical professionals and biologists on problems pertaining to spatial patterns of species in different remote areas, intra-year growth dynamics, GWAS, epidemiology and drug assessment. The posed problems called for a collection of statistical methods each time. The examples were generalized linear models, zero inflation, multivariate analysis of variance, survival analysis, latent state detection, propensity score matching, spatial statistics, ... Typically, genetics called for high-tech statistical weaponry while drug testing could be done using more traditional statistical tools. The following statistical packages were used in the process: R, Minitab, SAS, JMP, SPSS and Matlab. I have a PhD in Statistics (Berkeley).

Econometrics

Over the last eleven years I have worked with more than two hundred economists, social scientists, finance professionals and formally introduced finance companies, including hedge funds. We have dealt with problems pertaining to time series analysis, heteroskedastic models, panel data, structural equation modeling, robust inference, bootstrap, multi-factor representations, etc... Stata, R and EViews were my preferred tools for tackling econometric problems but I also gave clients the option to work in Matlab, SAS or Python. In my humble opinion, JMP, SPSS and Minitab are somewhat simplistic when it comes to serious econometric tasks. They are more suitable for other applied areas, like biology or quality control.