The great decrease in the cost of computing in recent years has made it possible to develop integrated analytical and statistical procedures which incorporate realistic biological structure for analyzing complex systems. This synthesis of biology and analysis holds the promise of producing solutions to many difficult problems in fisheries management. It is no longer necessary or even expedient to employ piecemeal methods developed for other purposes and adapted to fisheries problems. Instead it is possible to develop models of realistic complexity which incorporate robust estimation procedures which are specifically tailored to to the analysis of fisheries management problems.

Otter Research was formed by Dr. David Fournier and Dr. John Sibert to create software for rigorous statistical modeling of complex systems. MULTIFAN , Otter Research's length frequency analysis system, was released to the commercial market in 1988 and has users in many countries including Australia, Canada, Chile, France, Japan, Italy, Great Britain, Mexico, Venezuela, New Zealand, Norway, Senegal, Spain, Sri Lanka, Taiwan, and the United States.

Length frequency analysis refers to the analysis of collection of length measurements of fish sampled in some way from a fish population. For example, samples could be taken from a commercial fishery or from research cruises specially designed to sample young individuals. The purpose of the analysis is to derive estimates of parameters describing the population dynamics of the fish population which can form a basis upon which to formulate management policy.

MULTIFAN is a state-of-the-art integrated data analysis system for simultaneously analyzing multiple sets of fisheries length-frequency samples. The latest techniques for nonlinear statistical modeling and robust parameter estimation are exploited by MULTIFAN to provide reliable estimates of the parameters essential to length-based systems of fisheries stock assessment and management. A robust maximum likelihood method is used to estimate the proportions of fish at age in each sample and the parameters of the von Bertalanffy growth function. Additional options include estimation of seasonal variation in growth rate, total mortality, selectivity, and yield per recruit. MULTIFAN calculates approximate standard deviations and correlations of all parameter estimates. The use of a likelihood method enables the objective discrimination between models based on well established general theories of hypothesis testing using likelihood-ratio tests. MULTIFAN is designed for ease of use. All functions are accessed through a system of pull-down menus and data-entry windows. A comprehensive system of context sensitive on-line help screens provides both beginning users and sophisticated practitioners with detailed explanations of each step of the analysis. A system of automated tutorial demonstrations is available to instruct new users in the effective use of MULTIFAN.

MULTIFAN incorporates extensive graphics displays for initial data exploration and for viewing model results. Interactive graphics are used in setting up initial parameter estimates. Various text displays allow easy comparison of different models and statistical discrimination between alternative model hypotheses. Text screens may be copied either to a printer or to disk files for further processing by word processing systems or spreadsheets. Graphics and text screens may be printed on a dot matrix or laser printer.

MULTIFAN provides a simple collection of standard default menu options which will provide the correct model structure for analyzing most fisheries data sets. These defaults form a basis from which the user can easily develop more sophisticated model hypotheses for dealing with particular applications.

The user friendliness and flexibility of MULTIFAN encourage creativity in model formulation and data analysis, thus providing a powerful new tool for scientific data analysis and management.

figure 1

The application of length frequency analysis to a single sample may lead to ambiguous results. Simultaneous analysis of multiple samples, although much more difficult, eliminates these ambiguities. For instance consider the data presented in figure 1. There are three length frequency samples, each of which appears to have four modes. These data were derived by simulating a population with 7 age classes of which age classes 2, 4, and 6 are not represented in the samples. These data are intended to illustrate a common error which occurs when modes corresponding to dominant age classes are mistaken for modes corresponding to consecutive age classes. Two growth models appear to fit these data equally well when only the first sample is considered -- one with 4 age classes and K = 0.5 and one with 7 age classes and K = 0.25. The fits in figure 1 were derived by using parameters estimated by fitting the first length frequency data set alone. Although both growth curves fit the first data set equally well, it is apparent that the fish grow much more slowly than would be predicted by the 4 age class model with K = 0.5. This illustrates the extra information contained in a time series of length frequency data samples and the power of simultaneously analyzing the samples in such a time series.

MULTIFAN is tailored to your data analysis requirements with prices matched to the number of samples which MULTIFAN will analyze. Version 3.05 has the capacity to simultaneously analyze 5 samples and versions 3.10 and 3.15 will analyze 10 and 15 samples respectively.

Version Capacity Prices
3.05 5 samples $425.00
3.10 10 samples $800.00
3.15 15 samples $1025.00
32x 70 samples $1375.00
All prices are in US Dollars.
Visa and MasterCard accepted.
Prices subject to change without notice.

Multifan is available for 80386 and 80486 computers running MS-DOS 3.2 or higher. A graphics displays is essential and a math co-processor is required for 80386 machines. Approximately 570 Kb of free memory is required for MULTIFAN 3.15.

Multifan 32 is a protected mode 32-bit version of MULTIFAN using a 32 bit DOS extender requiring a 80386 (plus math coprocessor) or 80486 based microcomputer. It is compatible with DPMI32 and runs in a DOS window under windows 95 or NT.