MRP Group is a statistical consulting practice in Rochester, New York, led by Robert J. Parody, Ph.D. The person who scopes the engagement is the person who does the statistics and the person who defends them. Clients are organizations whose results have to survive a regulator, an auditor or a peer reviewer.
Robert J. Parody. Every engagement, start to finish.
Principal Statistician, MRP Group, Inc. · Associate Professor of Applied Statistics, Rochester Institute of Technology
Most consulting firms sell with their senior people and deliver with their junior ones. Here you work directly with the principal on every engagement — the same person in the scoping call, in the data, and on the line when a regulator or an auditor asks how a number was arrived at.
That direct involvement shows up in the small decisions as much as the headline ones. Which sampling scheme fits the process, whether a model assumption actually holds, which caveat belongs in the report and which is noise — these get decided by someone who heard the problem first-hand, not relayed through a summary. Questions get answered by the person who ran the analysis, and the reasoning behind a recommendation is always available rather than reconstructed.
The route matters, because it explains the range.
B.S. Chemical Engineering, Clarkson University (1998). Then four years in industry: Senior Quality Engineer at Alcan Aluminum and Celotex, facilitating FMEAs, control plans and measurement systems analysis under QS-9000, and completing Black Belt training through the AlliedSignal Operational Excellence program. Then Senior Engineer and Six Sigma lead at Pall Trinity Micro, launching the Six Sigma program at plant level and training Green Belts. The engineering degree is not decoration — it is why a conversation on a plant floor works, and why the statistics arrive in the vocabulary the people running the process already use.
M.S. Applied Statistics, Rochester Institute of Technology (2000). Ph.D. Statistics, University of South Carolina (2005).
The statistics degrees widened the work rather than replacing it. What followed was two decades across regulated domains that look unrelated from outside and are not:
Potency, stability and lot release; toxin neutralization assay ED50 estimation; lethal dose and exposure–response estimation by probit analysis across influenza, poliovirus and bacterial challenge models.
Study statistician across oncology, CNS and vaccine programs — protocols, statistical analysis plans, endpoint specification, CDISC SDTM and ADaM, and FDA Information Request responses on an active IND.
Six validation and reliability studies for a cognitive assessment device: test–retest reliability, inter-method agreement, covariate-adjusted normative ranges, and hardware and software release validation.
Authored statistical methodology content for the NUREG-1507 Section 9 revision for the Nuclear Regulatory Commission; guard band derivation from variance components for the EPA, delivered as regulator-risk and producer-risk curves.
Medicare Advantage utilization management and denials at state scale using CMS and all-payer claims data; national registry and survey analysis for HRSA and NSQIP-based surgical outcomes research.
The original thread, still running: process capability and control investigations, gauge R&R and measurement systems analysis, statistical process control and design of experiments.
Thirteen peer-reviewed publications span much the same range — quality engineering, hearing research, remote sensing, disaster risk, nutrition behavior and clinical outcomes. Breadth is not dilettantism here: the same small set of statistical ideas keeps turning out to be the answer in domains that have never spoken to each other.
Measurement, and what it can support.
The practice looks like three unrelated businesses from outside — federal regulatory work, industrial process and quality work, clinical and life sciences statistics. They are one business asking one question in three vocabularies.
Capability analysis, variance components and control charting — applied first to whether the measurement system can resolve the difference you are trying to control.
The question underneath every lot release decision. Control charting across a study series separates drift in the measurement from drift in the thing measured.
Guard bands, decision-error frameworks, and risk expressed on both sides of a limit rather than as a single threshold.
Test-retest reliability, inter-method agreement, covariate-adjusted normative ranges — a measurement-system question before it is a clinical one.
Sorting the handful of factors that genuinely move an outcome from the dozens that only appear to. Multivariate models, contribution decomposition and variance components, so effort goes where the variability actually is.
Confounding, collinearity and selection effects turn plausible associations into expensive mistakes. Establishing which relationships survive scrutiny is usually the difference between a finding and a decision.
The statistical machinery is the same across all of them. That is why a practice can credibly hold a medical device manufacturer, a federal regulator and a vaccine sponsor at the same time — and why the answer is sometimes that the data cannot support the decision being asked of it.
Associate Professor of Applied Statistics in the School of Mathematics and Statistics at Rochester Institute of Technology, since 2005. The courses are the practice areas: Statistical Quality Control, Design of Experiments, Applied Linear Models, Time Series Analysis and Forecasting, Categorical Data Analysis.
Two things follow from it. Methods get explained rather than asserted — a client should understand why an approach was chosen, not just receive its output. And the methodological base stays current by obligation rather than good intentions.
With a specific problem rather than a capability search. A gauge R&R that was rejected. A yield decline nobody can localize. An FDA information request that arrived with a clock on it. A study design that needs to survive review before enrollment opens.
The first conversation is diagnostic and free: what the decision is, what data exists, and whether statistics can actually get you there. Sometimes the honest answer is that the design cannot answer the question and no analysis will rescue it. Being told that in week one is worth more than being told it in month four.
Robert J. Parody, Ph.D. Every engagement is scoped, led and delivered by the principal, so clients work directly with the statistician doing the analysis — from the first scoping conversation through to defending the method to a regulator, an auditor or a peer reviewer.
Measurement. Whether a process has lost capability, whether an assay is drifting or the product is, whether a device agrees with its reference instrument, and whether a single passing test proves compliance are all the same question in different vocabularies: what can this measurement actually support? The statistical machinery — variance components, capability indices, control charting, agreement analysis — transfers directly between them.
MRP Group, Inc. is a New York business corporation, filed 28 April 2022, DOS ID 6470151, based in Rochester, New York. It is a small business with no socioeconomic set-aside designations. UEI RPKJGEB3QKL8, CAGE 9AWW3. The practice previously operated under the name Responsum Analytics.
No. All federal past performance is as a subcontractor, through prime contractors. MRP Group holds no prime federal awards and claims none. The firm holds GSA Multiple Award Schedule contract 47QRAA25D000P under SINs 541990 and 54151S, which lets a prime add the awarded Statistician labor category to a team without a rate crosswalk.
