Statistical Data Analysis Training
Statistical Data Analysis Training Course Description
This course is valuable for anyone who finds a need to understand or apply statistics. The course can serve as a refresher or as a practical introduction. The topics presented were selected based on Dr. Olsen’s experience working for over 20 years on Navy applications. The statistical techniques often used by weapon system analyst are carefully explained. The presentations emphasize the intuitive development and the practical use of the techniques rather than providing academic developments. Weapon System and other examples are used to illustrate and motivate the topics. The course provides a “toolbox” of statistical methodology and the necessary understanding teaches the student to avoid and detect the common mistakes encountered in applying statistics.
With onsite Training, courses can be scheduled on a date that is convenient for you, and because they can be scheduled at your location, you don’t incur travel costs and students won’t be away from home. Onsite classes can also be tailored to meet your needs. You might shorten a 5-day class into a 3-day class, or combine portions of several related courses into a single course, or have the instructor vary the emphasis of topics depending on your staff’s and site’s requirements.
Introduction—A review of the basics, clarifying notation. Topics include conditional probability and the low base-rate problem, simple probability models, expected values, and how to relate measurements using joint distributions, correlation’s and covariance’s.
Reliability Models and Decision-Making—Discrete probability models for use in reliability and other analysis are introduced. False alarm rates, statistical power, P-values, and (R) OC curves commonly seen in computer output are explained. You will master step-by-step procedures for solving problems.
Decisions Making Based on Continuous Measurements—Continuous probability models, including the normal, the chi-square and the student “t” are introduced. These models are used for both weapon system accuracy and reliability. You will understand why these models are often used and when to apply them. False alarm rates, statistical power, P-values, and (R) OC curves are explained for data from the normal distribution. An understanding of these topics is necessary for the proper use of statistical software. These topics also provide a fundamental understanding necessary for sample size determination.
Confidence Intervals—An analysis of test results requires an understanding of decision making and the use of confidence intervals. You will be able to determine sample size requirements and will be able to understand and present results in different ways. Exact methods for placing confidence intervals on system reliability are presented. These methods are particular useful for highly reliable systems.
Comparing Results—Comparing the performance of two systems requires specialized statistical methods that make different assumptions about the nature of the application. You may want to compare rocket motors manufactured by two different venders or to compare a new system to an old one. You will learn to use techniques for small and large samples and paired observations and be able to choose between the different analysis techniques.
Linear Regression—You will learn how to use regression to build and investigate models. Regression is often used to model system performance as a function of controlled or environmental variables.
Multiple Regression—You will learn how to build complex models for prediction results and for identifying possible cause and effect relationships. Interaction terms, indicator variables and polynomial models will be explained. You will also learn how to evaluate relationships between measurements both in statistical terms and in terms of total uncertainty. Topics include R-square, partial correlation coefficients, hypotheses testing and analysis of variance.
Logistic Regression—This wildly applicable and powerful analysis technique is rarely taught outside of specialized courses because it is relatively new and it requires statistical software. Since it is important in weapon system analysis, including the detection of reliability changes with time and in understanding the effects of controllable variables, it is being presented in this course. The technique is simpler to use than neural networks and often provides better solutions.
Test for Normality—Many analysis techniques assume that the data are normally distributed. Impact data provides such an example. You will learn to test the assumption of normality.
Whether you are looking for general information or have a specific question, we want to help
Request More Information