FPL Statistics Unit

Statistics Unit's FY98 Attainment Report

A. Partnerships

1. With other FS statisticians

As a result of the National Workshop on Forest Service Research Statistical Functions held in September 1997 in Washington, D.C. with Rich Guldin cooperation with other Forest Service statisticians continues developing on several levels. The statisticians have been asked to consider ways to evaluate the statistical quality of research activities at each Station. Summaries of the statistical activites at each Station have been requested, as has the development of a nation-wide contact booklet. Statistical units are to consider the value of a charter or similar type document for each Station. The statisticians are asked to increase interaction with ARS statisticians. To promote the use of study files in Forest Service research, we should evaluate the use that we might make of RBAIS to serve as an electronic study file. We should continue to work together in an effort to supply needed statistical software to FS personnel. This effort results from an effort initiated on February 8, 1994, by William Bristow, Chief Information Officer for the Forest Service when he asked a committee to work on how best to meet the needs of Forest Service employees in a comprehensive manner. However, we need to phase out support to the Regions if there is no effort by the Regions to help develop a more permanent solution. James Evans will have primary responsibility for each of the preceeding activities.

Finally, we need to work with the other Statistics Units to link resources on the World Wide Web. This effort is aimed at coordinating and making more efficient each unit's efforts to provide needed statistical information to FS scientists. This effort will provide ways of sharing educational material on how to do certain statistical procedures, it will share research results and it will provide a technology transfer opportunity for statistical software we produce. By doing it on the WWW, we also open up opportunities to reach an audience broader than just the FS. We are asked to develop a proposal for this statisticians web page and submit it to the SPPII staff for sponsorship. Most of the work this year will be by Steve Verrill, Vicki Herian and James Evans.


--To develop a recommendation on how we can evaluate the statistical quality of the research activities at each Station.

--To create a summary of the statistical activities at each Station and to develop a nation-wide contact booklet.

--Completion of the RWUD to serve as the charter for FPL's statistics unit.

--Work to develop closure on the efforts to provide statistical software to Forest Service personnel.

--Look for ways to interact with ARS statisticians

--Evaluate and make recommendations as to how RBAIS can be used to promote the concept of study files.

--Work with other station statisticians to develop a Forest Service Statistics home page on the WWW that can provide information on the statistical function in the FS, provide training on the use of statistics, and serve as a technology transfer effort for statistical developments.

--Continue the development of an interactive set of computer programs that will run across the WWW to explore the uses of the WWW in technology transfer and pilot opportunities for the other Statistical Units and other FPL units.


--A draft document called "Status of the Statistical Function in Forest Service Research and Development" was prepared and sent to Rich Guldin. This document proposes a national assessment of both the statistical functions and of the statistics being performed outside the statistics units.

--A nation-wide contact book for statistics was developed and a summary prepared on the statistical activities at each Station and FPL. These were sent to Rich Guldin

--Completed RWUD to serve as charter for FPL's statistics unit

--Working with other station statisticians to define and create a Forest Service Statistics home page.

--Steve Verrill and Vicki Herian made invited presentations to the 1998 Forest Service Research Statisticians' meeting in Dallas on how to set up interactive computer programs on the WWW.

--Helped develop a joint review pilot program with other FS Research Statisticians

2. With other FPL scientists and their outside cooperators

Through their statistical consulting activities, members of the Statistics Group are actively involved in many of the partnerships that FPL scientists are developing. All members of the Statistics unit are involved in these efforts.


--Emphasize service to projects with outside cooperators, giving special effort to produce quality support and still meet the deadlines of these projects.

--Analyze data on using image analysis for flake furnish characterization in coop study of Bob Geimer with Washington State University, University of Tennessee, and University of Maine.


--Completed analysis of data in flake furnish characterization study of Bob Geimer. Draft paper written. Work an reanalyzing a subset of the data is in progress

--Assisted Gary Gramp (Technical Director, HPVA) and Gary Heroux (Director, Laboratory Services, CPA) with the statistical aspects of a paper.

--Helped design study and did preliminary analysis on Bob Falk's Fort Ord study on recycled lumber. This is a cooperative study with Fort Ord Reuse Authority and West Coast Lumber Inspection Bureau.

3. With standards and regulatory committees

The Statistics Group is regularly involved in submissions to the Board of Review of the ALSC such as expected submissions of Swedish Spruce and Austrian Spruce this year. The unit is also actively working with FPL scientists on consensus standards committees like ASTM and AWPA. Some of these efforts include standards on monitoring changes in lumber properties and standards on evaluating the performance of preservatives. Primary responsibility for this in the unit is with the senior staff.


--Help FPL meet its advisory commitments to groups like the BOR of the ALSC and to help FPL scientists in developing and getting standards approved.

--Provide support to FPL involvement with consensus standards committees like ASTM and AWPA.

--Evaluate modified grouping criteria for D1990

--Review Canadian methodology for their ALSC submission to see how it might differ from other species

--Analysis of Englemann Spruce and Balsam Fir from Minocqua mill


--Reviewed the Canadian methodology for their submission of In-Grade data to the BOR of the ALSC and reported or results.

--Provided a technical review of a submission on design numbers for Austrian Scots Pine and convinced WCLIB to use cell by cell adjustments due to failure to meet the GQI test in ASTM D1990.

--Reviewed the method Al Debonis is using to develop strength ratios of a Lithuanian Spruce submission and corrected his procedures

--Worked with Stan Lebow and Jerry Winandy to develop an AWPA fixation standard for preservatives

--Developed and documented a Web-based program for Brad Douglas and Bob Tichy that permits them to evaluate criteria for acceptance testing of duration of load samples.

--Modified the Web interface of the nonparametric estimation program to make it recognize and handle blank lines at the end of a data set submitted to the BOR as support of a submission. This was needed to help Al Debonis in his preparation of a submission on Lithuanian Spruce.

--Created a Web program that takes the final characteristic value for MOR and using ASTM D1990 procedures calculates unrounded design values for MOR, UTS and UCS. This program was used by Al Debonis in his submission on Lithuanian Pine.

--Completed reviews on Austrian Scots Pine, Lithuanian Spruce and Austrian Spruce submissions to the BOR of the ALSC with recommendation to accept the first two and reject the last one until corrections are made. These recommendations were accepted by the BOR

--Began work late in the FY on four submisions to the BOR on Lithuanian Pine, Swedish Spruce, a revised Austrian Scots Pine and a revised Austrian Spruce submission.

B. New studies/work

1. Concommitance Research

To adequately model the performance of a wood member under combined stresses like bending and compression, it is necessary to know the relationship between the strength properties of the wood member. The FPL Statistics Unit created methods for estimating the correlation between strength properties that can each only be measured through a destructive test. This work, published over 10 years ago, has been a source of the development of several competing methods in the statistical literature. It is time to do a comparison of different methods available and look at what impact they might have on estimates from FPL studies. In addition, we need to extend the procedures beyond the assumption of bivariate normality to include bivariate distributions whose marginal distributions may better represent the univariate distributions of lumber strength properties. Primary responsibility for this research will be James Evans, Patti Lebow and Dave Green.


--Run a comparative study of existing methods of estimation of the correlation between lumber strength properties and look at its implication to FPL modeling efforts.

--Extend the concept of concommitance to bivariate distributions other than the bivariate normal and begin studies on how to estimate relationships


--On hold. This area of research has been replaced by a study on Weibull distributions because of a chance to write a book chapter on the use of Weibull distributions in wood engineering. This opportunity has allowed us the chance to develop publications from previous research we have done and to fill in gaps with new research. Two publications were AD approved this FY and two more are in progress.

2. Development of Economic Models Use Distributional Information

The NASAW and NAPAP economic models rely on the PELPS computer program for making important economic projections concerning the US and Canadian solid wood, and pulp and paper sectors, respectively. It is desired to have a better understanding of the PELPS source code in order to update the code as necessary to enhance the NASAW and NAPAP models. This should allow us to look at incorporating coefficient distributional information into the model which could allow us to produce some type of confidence limits on the estimates the model produces. This work will be joint with Patti Lebow and members of the Economics RWU.


--Develop a user manual and written documentation of the code description

--Develop a study plan on how we could add distributional information to the models and how we could evaluate the results of our confidence estimates

--Make the models more Windows 95 friendly


--User manual is in progress

--The PELPS system was updated to run in a console (DOS) window under Windows 95 along with 123-97. Progress is being made on a Windows 95 graphical interface.

3. Moving information off the DG Computers and the UW Computers

During the next year, FPL will be moving from the DG computer system to an IBM Unix Server computer system. At the same time computer tapes stored at UW and used to store historical work and data sets that we periodically need to reference, have passed the life expectancy of such tapes. The Statistics Unit has numerous files that will need to be evaluated and moved to either the IBM system, our optical drive, or both. Failure to do this could cause a loss of critical material. This will involve months of effort. Primary responsibility for this will be with James Evans and Cherilyn Hatfield.


--To move all information that we need to save from the DG and University computer systems to either the IBM system or our optical drive.


--All of the the Statistics Unit's data at UW and on the DG has been archived to the optical disk, including 60 years of FPL-02 historical data.

--All important documents on the DG computer were moved to the new IBM system by September 1, 1998. The unit is totally over and using the new system for everything except for processes such as APOS that have not yet been converted to the IBM system.

--We have also helped move information from UW tapes to the optical drive for RWU 4714 and RWU 4723.

C. Continuing studies

1. Support to FPL's QA/QC program, Ecosystem Management Research and Composite Research

To be a world leader in research on wood utilization, it is important that the Forest Products Lab do the right research and do it well. Members of the Statistics Unit are involved in helping FPL identify what research should be done through service on the Ecosystem Management Steering Committee and the Composites Research Steering Committee. These committees help identify high priority research opportunities and help integrate these opportunities into the FPL research program. Helping FPL be a world leader and a standard of excellence for quality and credibility in natural resource research is a key role of the QA/QC Steering Committee. Members of the Statistics Unit will play a critical role on this committee in its initial year of effort.


--Help implement a QA/QC program at FPL through service on the FPL QA/QC Steering Committee

--Help identify new opportunities for FPL research through service on research steering committees.


--QA/QC committee prepared FPl's response to the WO's request for an implementation plan. This involved having an all scientist meeting to gain support and acceptance for the proposed response. The committee also created the plan for rotation of members, with a new committee being appointed for FY99.

--The composite research stearing committee is putting together a research plan that includes our current and planned future research. This plan will be reviewed by a industry advisory committee in June. The Statistics unit is active on two of the major research areas and supports FPL scientists in all the research areas.

--Created one page sheets outlining needed information in study plans and study files

2. Lumber Property Studies

Numerous factors such as temperature, moisture content, knots, grading methods, and others effect the properties of structural lumber. The Statistics Unit continues cooperation with RWU 4714 on several studies. The In-Grade Testing Program is one area of this cooperation. It is a cooperative program between FPL, representative lumber associations, and faculty at several universities. Extensive work has been done in the program that has resulted in ASTM standards for obtaining design values from In-Grade Testing and with revised lumber properties for the major commercial US species. The additional work in this program centers on developing a series of research papers on the lumber properties using more current adjustment models for environmental effects such as moisture content. This work will require re-analysis of the In-Grade data using revised methodology and will study the implications of information learned about the species. Modeling the effect of environmental factors on lumber properties will also be done. Other areas of cooperation with RWU 4714 includes studying lumber grading criteria, use of new wood species, and lumber sorting procedures such as log thumping. This research is a collaborative effort with Dave Green and Dave Kretschmann of RWU 4714 and James Evans and Cherilyn Hatfield from the Statistics Group.


--Reanalysis of Southern Pine data using FPL's recommended procedures.

--Paper on the effect of temperature on lumber properties

--Paper on DF log thumping study to look at effectiveness of prediciting board properties from log properties

--Analysis of Alaska Yellow Cedar bending data

--Analysis of static short term properties for temperature data

--Evaluation of effectiveness of sorting SP by log thumping

--Stress wave analysis of long term temperature study


--We have analyzed new data on the short term effect of temperature on lumber properties and worked with Dave Green to develop models and prepare a presentation that was given in Sweden and at the NDT Symposium in Madison.

--Other studies are on hold

3. Modeling the effect of moisture content on lumber strength properties

Extensive efforts at modeling the effect of moisture content on lumber strength properties have resulted in the development and study of more than one hundred models of numerous forms. Currently models for adjusting lumber strength properties of MOR, MOE, UTS and UCS have resulted from this research and are incorporated in ASTM standards. However, this research was done on only two species of lumber, Douglas Fir and Southern Pine, both big knotted species. Current research centers on the development of models for a wide range of species. This research is joint with Dave Green and James Evans.


--Create User Manual for a WWW interactive program that allows a user to predict changes in lumber strength properties from changes in moisture content. Program allows adjustment for bending, tension, and compression strength as well as modulus of elasticity using both ASTM models and FPL recommended research models.

--Analysis of the moisture content hysteresis effect on tension properties of Southern Pine

--Paper on modeling MC adjustments


--On hold

4. Development of statistical tools for FS scientists

Work on the In-Grade program and other research programs have resulted in the partial development of many new statistical procedures and tools. This research is aimed at completion of earlier work to produce usable statistical tools for FS researchers and to provide these tools in a form that is readily available to FS researchers. Current efforts deal with developing technology transfer computer programs that allow researchers to estimate lumber strength properties based on distributional percentiles. Nonparametric estimates are currently being incorporated into user friendly programs. Programs that give researchers the tools to perform Weibull, normal or lognormal estimates of properties and confidence intervals with guidance for choice of distributional form are being developed. This includes development of new statistical tests for evaluation of distributional fits. Research on predicting the correlation between strength properties that can only be measured through destructive testing is being extended to a different distributional forms to allow researchers to better choose relationships between properties. This research involves in various combinations Dave Green from RWU 4714, James Evans, Steve Verrill and Vicki Herian.


--Revise WWW and floppy disk versions of a computer program that allows scientists to get nonparametric percentile estimates and tolerance limits from a data set and allows scientists to know what order statistics relate to different percentile and tolerance limit estimates for different sample sizes. Also develop draft users guide for the program

--WWW and floppy disk versions of a computer program that allows scientists to fit a two or three parameter Weibull distribution to a data set and get goodness-of-fit tests, percentile estimates and tolerance limits.


--A program that allows the user to fit normal, lognormal, two-parameter Weibull and three-parameter Weibull has been written and is being tested before putting up on the WWW.

--As part of our development of statistical routines that can be run over the WWW, translated LINPACK linear algebra routines and UNCMIN nonlinear optimization routines into Java and made source code available on the Web. This code has been downloaded by 2000 users.

5. FRT Serviceability Project

This study is to address concerns of fire retardant treated plywood and lumber used in roofs and includes the study of kinetic models to explain mechanisms involved in predicting end-use performance. Part of this research is conducted in cooperation with the State of New Jersey. The research team consists of several FPL scientists and is led by Jerry Winandy. From the Statistics Group the major involvement is by Patti Lebow.


--Work with team to develop RBD serviceability model.

--Develop empirical-mechanistic model of hemicellulose concentrations in FRT degraded clearwood which can be linked to strength degradation.

--Verify previously developed kinetic model of FRT degraded clearwood by addition of 3 and 4 year data.


--Previously developed kinetic models were verified; a paper summarizing this work was accepted by Wood & Fiber Science

--Progress continues on the chemical composition relationship to FRT strength degradation. More data has become available and we are re-analyzing the acid systems

6. Analyzing Underdetermined Data

In recent years "chemometricians" have developed a tool, partial least squares (PLS), for analyzing underdetermined data. (Data is underdetermined if there are more predictors than observations. For example, if one prepared 7 paper specimens and then measured 30 strength predictors for each of the 7 specimens, one would have an underdetermined data set. It would not be possible to obtain a unique least squares fit that predicted strength as a function of the 30 predictors from this data set.) PLS has been gaining acceptance among chemists. However, there are competing technologies; among them, principal component regression (PCR), ridge regression (RR), and variable subset selection (VSS). With Umesh Agarwal we intend to investigate the merits of these competing technologies as applied to Raman spectoscopy. As part of this work, we also expect to develop user friendly, public domain computer routines that implement the most promising of these techniques. This work is being done by Steve Verrill and Umesh Agarwal.


--Evaluate the technology on a broad range of data sets

--Polish computer routines that implement the technology


--Programs have been improved and applied to expanded data sets. Research and development work in this area continues

7. Monitoring lumber properties

Extensive effort went into evaluating lumber strength properties for dimension lumber as part of the US Ingrade Program. Now that properties have been developed and approved by the BOR of the ALSC, there is a need to develop standards on how to monitor potential changes in these properties over time. Numerous possible procedures are available to do this. However the sensitivity of these procedures to actual changes in properties, and to expensive retesting whan not needed, must be evaluated in order to produce a cost effective procedure. Primary responsibility for this work will belong to Dave Kretschmann and James Evans.


--Write paper on results of a simulation study on possible monitoring procedures


--Draft paper written and out for review

8. Statistical Issues in Preservative Testing

Preservative testing is aimed at answering questions like "Will a preservatively treated product meet some specified lifetime consistently?", or "What is the degredation process?', or "How does it compare to current systems in similar use situations?" These questions have highly important statistical issues associated with them. The Statistics Unit has a long history of supporting FPL research on preservatives on individual studies using whatever personnel were available at the time. However there is an opportunity to develop a more integrated statistical program with the preservative testing program. Attempts have been made in the past when collaboration between Statistics and the Preservative Testing Team produced memoranda on "Experimental Design Issues for the Studies on Efficacy of Preservative Treatments" and on "Statistical Approaches to Looking at Stake Testing." These memoranda primarily identified major statistical issues that affect our capabilities in preservative testing. They did not offer solutions. Now is the time for more coordinated support to FPL's preservative testing program that incoporates both the Statistics Unit's capabilities in providing statistical research support and the Statistical Unit's data handling and analysis support. Primary responsibility for this effort will be with Rod DeGroot, Doug Crawford, Stan Lebow, Patti Lebow, Steve Verrill, Cherilyn Hatfield and James Evans.


--Complete computer program that helps update stake testing data sets and provides in the field analysis capabilities

--Provide support to FPL efforts to evaluate the copper tolerance of various fungi.

--Provide support to FPL efforts to study the durability of Alaska Yellow cedar

--Provide design and analysis of current studies on leaching rates


--Three studies on copper tolerance of various fungi have been analyzed with information sent back to the scientist.

--The Fortran program for stake testing has been updated for output of the maps of the field plots. The SAS program has been rewritten and a field program is in development.

--Analysis of leaching larger-size CCA treated specimens in fresh and sea water is complete with a poster presented at the AWPA meeting in May and a paper AD approved and nearly ready for submission to the journal.

--Analyzed Alaska yellow cedar decay data for Rod DeGroot

9. Microfibrial angle

It is believed that microfibrillar angle is closely related to wood strength properties. However it has been difficult to investigate this relationship due to the laborious process (a manual one involving dyed specimens and microscopy) needed to obtain the microfibrillar angle measurements. Recently scientists have been investigating the feasibility of replacing these manual methods with computerized x-ray diffraction techniques. Harry Alden has made manual measurements on a large collection of specimens. We will make use of the University's Hi-STAR x-ray diffraction equipment and computer routines developed by the Statistics Group to investigate the relationship between manual and computerized microfibrillar angle measurements in this collection. This work is joint with Steve Verrill, Harry Alden and Dave Kretschmann.


--Prepare paper on results


--Paper presented at international comference. This paper was published in the preceedings of the conference.

--A grant for continuing this work was prepared and approved

[Forest Service] [Forest Products Lab] [FPL Statistics Unit]

We welcome your comments and suggestions.

This page was last modified on January 22, 1999.