SRL Quantitative specializes in R and C++ for statistical programming. Beacause R interfacess seamlessly with C++, this combination offers the perfect balance of ease of use and efficient calculation without the worry of licensing costs and subscriptions. R provides an ideal platform for data management and workflows while C++ offers faster performance than anything else on the market, open source or otherwise, allowing us to use larger and more sophisticated models than other software can handle.
Our focus is on predictive statistics for time series data. You can access our open source Bayesian Dynamic Factor Model R package on GitHub, or contact us for models that are tailored to your own needs, including mixed frequency modeling, targeted predictive models (i.e. focused on predicting a single series), or forecast breakdowns.
Good predictive statistics can help your business reach its goals, but forecasts for large data sets can be overwhelming. We can help you make the most of backcasts, nowcasts, and forecasts by calculating the optimal response to quantifiable goals. In particular, we can help you develop cutting edge portfolio optimization routines for quantitative finance.
We write the majority of our code in R and C++ to offer an optimal combination of usability and speed. Our open source code, available on GitHub, offers unparalleled efficiency and functionality. You can estimate Bayesian posterior densities for dynamic factor models in seconds, or estimate likelihood based seasonal adjustments using exogenous factors. For Bayesian routines, saved output includes entire posterior distributions, not just point estimates. If you want to go beyond our open source code, we can tailor software to your needs. Because our predictive statistics are fully Bayesian, custom routines can produce the full posterior distribution of pooled forecasts, maximizing information content. And because we do not use black-box methods, you will always know what is driving your predictions. Depending on your needs, we can provide software in R and C++ or stand-alone applications that do not require any coding skills at all.
Our books are availabe for free download: Practical Implementation of Dynamic Factor Models and Tools for Macroeconomic Theory. You are welcome to use these texts in your studies, or to work through them with your industry team. We can also help you generate technical documentation for your own software or statistical methodology for either in-house use of provision to clients.
Practical Implementation of Dynamic Factor Models is exactly what the title states: a practical guide to implementing dynamic factor models. However, it also includes many of the necessary derivations including posterior distributions for parameter estimates using normal-inverse gamma and normal inverse-Wishart conjugate priors. The book is geared towards forecasting applications and the notation follows the syntax of our open source R and C++ code, which we hope will make these programs easier to follow.
Tools for Macroeconomic Theory: Local Solution Techniques introduces the reader to finding log linear solutions to macroeconomic models and stochastic systems of difference equations more generally. Chapter 1 introduces Lagrange's method for constrained optimization illustrating why Lagrange's method works and how we can interpret the Lagrange multiplier. Chapter 2 introduces dynamic programming and offers examples of solving both unconstrained and constrained optimization problems. Chapter 3 introduces the Hamiltonian, the continuous time equivalent to the Lagrangian and concludes the first step of solving macroeconomic models: finding the first order conditions for an optimum. The second step in solving macroeconomic models by hand is finding the deterministic steady state of the model and approximating the model around this point; this is the subject of Chapter 4. Chapter 5 covers the final step in solving macroeconomic models with pen and paper: solving the stochastic system of difference equations described by the log linear model. This is typically regarded as the hardest part of traditional macro and to date one of the best references remains Blanchard and Kahn's 1980 article. This book offers a simplified solution technique making it accessible to the beginning economics student. The chapter ends by presenting results for several models that I develop throughout the proceeding chapters. Finally, Chapter 6 works through a reasonably simple model from start to finish.
Get the most out of your data with seamless transitions from raw data to predictive modeling and quantifiable goals. While our C++ software can do some heavy lifting, our workflows have a strong emphasis on simplicity. And because everything we provide is in R and C++, there are no subscription fees and you will have access to a vast on-line support community. Get in touch to inquire whether our solutions may be right for your business.
Use the power of R to make your data management simple, efficient, and compatible with almost any other software. Go effortlessly from any format to matrix layout for statistical applications, select observations based on their attributes, or add attributes to existing data. Our specialty is in time series data.
Bring Bayesian statistics to your data. Our time series routines focus on backcasting (predicting past data that has not yet been released), nowcasting (predicting current data that has not yet been released), and forecasting. Depending on your objectives we can help you develop predictive statistics for individual series or for a whole panel of data. And because we don’t use black-box or discretionary methods, all of our routines can be rigorously back-tested prior to implementation.
For predictions of large sets of data --- with hundreds or even thousands of series --- the optimal response based on forecasts is not always clear. We can assist your decision-making framework by quantifying your objectives and solving for the best course of action. These problems can be unconstrained --- in which we look for a global optimum --- or we can restrict input variables such as budget or portfolio size to a fixed amount. Whenever possible we’ll use matrix calculus to find closed form solutions. Even when we are forced to use numerical solution techniques (which can get stuck in local optimums), we’ll begin with an approximate closed form solution to make sure we are working around the best possible outcome.
Let us help get your team up to speed using R for data management and predictive modeling, in using our open source software, or in Bayesian statistics more generally. Courses are tailored to your needs, and can be used to develop software for your business using your data. We can come to you or you can join us where we are based on the Big Island of Hawai’i.
SRL Quantitative is part of SRL Analytics, a sole proprietor LLC founded by Seton (Seth) Leonard in 2016 as a statistical programming consultancy. Seth collaborates with other programmers and consultants on an individual project basis. He earned his Ph.D. in economics in Geneva, Switzerland in February 2017. The company is based on the Big Island of Hawai’i.
+1 (603) 229-2095