Statistical Modeling

We can help to formulate, solve, and promote dissimilar business problems in finance, marketing, retail, health care, real estate, manufacturing, transportation and other industries that require the usage of quantitative methods, such as:

  • Study Design

  • Survey Sampling

  • Exploratory Data Analysis and Multidimensional Data Visualization

  • Any type of Regression Modeling (linear and non-linear, parametric and non-parametric, with binary, categorical, ordinal, count or continuous dependent variable, for correlated and uncorrelated data)

  • General linear and non-linear regression, Generalized linear and non-linear regression

  • Linear and non-linear regression with any types of constraints

  • Regularized regression (Generalized Path Seeker regression, including Ridge regression and LASSO)

  • Generalized additive models and Spline regression

  • Benchmark regression (Stochastic Frontier Analysis)

  • Switching regression

  • Robust regression

  • Quantile regression

  • Bayesian regression

  • Structural Equation Modeling

  • Finite Mixture models

  • Latent Class Models

  • Markov Chain Models

  • Analysis of Survival Data

  • Statistical Quality Control

  • Multivariate Analysis (cluster analysis, discriminant analysis, factor analysis, principal component analysis, partial least square regression, biplots, etc.)

  • Econometrics, Time Series Analysis (univariate and multivariate) and Forecasting

  • Spatial Statistics and GIS

Corresponding data can be:

  • primary or/and secondary

  • cross-sectional, time series, panel or cross-sectional time series

  • spatial

  • transactional

We are experts in the analysis of small data, data for which the number of predictors is larger than the number of observations, and in incorporating expert judgement in the modeling of such data. In particular, the regression analysis of data for which the number of observations is smaller than the number of predictors can be done via Generalized Path Seeker regression, LASSO regression, and Support Vector Machine regression.

The regression analysis of small data is meaningful even for only five observations. An adequate approach is a Generalized Maximum Entropy regression with priors. Priors can be based on expert judgement and can be set up as histograms or as intervals for unknown parameters. Another approach to consider for small data analysis is Bayesian regression.

Project Assessment

Project assessment starts with the translation of a business problem into its solvable counterpart and evaluation of the feasibility and success of a project implementation. 

In particular, the project assessment phase includes:

  • evaluating business objectives and goal attainability

  • checking the correspondence between objectives (business questions) and data/information availability and data sufficiency

  • discussing variety of problem statements, statistical hypothesis, dependent and independent variables and selection the most appropriate ones

  • discussing uncertainty in the problem parameters, data, project duration, project schedule risk analysis (if necessary), requirements, and outcomes

  • determining required resources

  • discussing integration of the solution of the problem  with your planning / forecasting / decision support systems and post-solution activities

  • developing customized report structure

  • forming a multidisciplinary/cross-functional project team (if necessary)

Combining statistical modeling with data mining/machine learning, optimizationforecasting, and spatial analytics can significantly widen the list of problems, mentioned above, and lead to additional increments in ROI and other benefits.


A typical statistical project includes the following steps: 1. project assessment, 2. detailed formulation of a statistical analysis/statistical modeling problem and methodology development, 3. finding corresponding solutions, 4. deployment of statistical application within the company's decision support system.

Detailed formulation of a statistical problem

Reviewing available data/information sources in light of project objectives and business questions, collecting necessary expert judgments and creating data set for statistical analysis. Defining variables' role and formulating the business problem in statistical terms. When available data is insufficient, or the requirement is to develop an "if-then" scenario, then an appropriate approach can be simulation. Developing an adequate methodology/selecting an adequate statistical method to solve the problem at hand. Calculating a justifiable sample size (utilizing power analyses).

Finding an acceptable solution

Converting formulated statistical problem into solvable counterpart by applying an available commercial tool (SAS, EViews, SPSS, Latent Gold, etc). Developing the application to produce the solution, generating the solution and analyzing its properties. Interpreting and presenting results.


Deployment of the application within the company's decision support system

Assessing model's (findings) recommendations. Creating software infrastructure to incorporate model/findings into business decision making. Setting up the report structure and reporting system. Measuring performance accuracy, identifying factors that drive discrepancy between reality and expected outcomes, and modifying/updating data/model. Embedding statistical application into company's decision support system.