Designation: D 2915 – 03

Standard Practice for

Evaluating Allowable Properties for Grades of Structural
Lumber1
This standard is issued under the fixed designation D 2915; the number immediately following the designation indicates the year of
original adoption or, in the case of revision, the year of last revision. A number in parentheses indicates the year of last reapproval. A
superscript epsilon (e) indicates an editorial change since the last revision or reapproval.

INTRODUCTION

The mechanical properties of structural lumber depend upon natural growth characteristics and
manufacturing practices. Several procedures can be used to sort lumber into property classes or stress
grades, the most widely used being the visual methods outlined in Practice D 245. With each, a
modulus of elasticity and a set of from one to five allowable stresses may be associated with each
stress grade. The allowable stresses are extreme fiber stress in bending, tension parallel to the grain,
compression parallel to the grain, shear, and compression perpendicular to the grain. This test method
for evaluation of the properties of structural lumber defines an allowable property as the value of the
property that would normally be published with the grade description.
This practice is useful in assessing the appropriateness of the assigned properties and for checking
the effectiveness of grading procedures.
For situations where a manufactured product is sampled repeatedly or lot sizes are small, alternative
test methods as described in Ref (1)2 may be more applicable.
1.2 The values stated in inch-pound units are to be regarded
as the standard.
1.3 This standard does not purport to address all of the
safety concerns, if any, associated with its use. It is the
responsibility of the user of this standard to establish appropriate safety and health practices and determine the applicability of regulatory limitations prior to use.

1. Scope
1.1 This practice covers sampling and analysis procedures
for the investigation of specified populations of stress-graded
structural lumber. Depending on the interest of the user, the
population from which samples are taken may range from the
lumber from a specific mill to all the lumber produced in a
particular grade from a particular geographic area, during some
specified interval of time. This practice generally assumes that
the population is sufficiently large so that, for sampling
purposes, it may be considered infinite. Where this assumption
is inadequate, that is, the population is assumed finite, many of
the provisions of this practice may be employed but the
sampling and analysis procedure must be designed to reflect a
finite population. The statistical techniques embodied in this
practice provide procedures to summarize data so that logical
judgments can be made. This practice does not specify the
action to be taken after the results have been analyzed. The
action to be taken depends on the particular requirements of the
user of the product.

2. Referenced Documents
2.1 ASTM Standards:
D 198 Test Methods of Static Tests of Timber in Structural
Sizes3
D 245 Practice for Establishing Structural Grades and Related Allowable Properties for Visually Graded Lumber3
D 1990 Practice for Establishing Allowable Properties for
Visually-Graded Dimension Lumber from In-Grade Tests
of Full-Size Specimens3
E 105 Practice for Probability Sampling of Materials4
3. Statistical Methodology
3.1 Two general analysis procedures are described under
this practice, parametric and nonparametric. The parametric
approach assumes a known distribution of the underlying
population, an assumption which, if incorrect, may lead to

1
This practice is under the jurisdiction of ASTM Committee D07 on Wood and
is the direct responsibility of Subcommittee D07.02 on Lumber and Engineered
Wood Products.
Current edition approved April 10, 2003. Published June 2003. Originally
approved in 1970 as D 2915 – 70 T. Last previous edition approved in 2002 as
D 2915 – 02.
2
The boldface numbers in parentheses refer to the list of references at the end of
this practice.

3
4

Annual Book of ASTM Standards, Vol 04.10.
Annual Book of ASTM Standards, Vol 14.02.

Copyright © ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959, United States.

1

). with the use of the following equation.2 Where possible.2 Sampling with Unequal Probabilities—Under some circumstances.D 2915 – 03 may be estimated. the desired content (C) (Note 3) and associated confidence level must be selected. For all properties. For the five allowable stresses and the modulus of elasticity various percentiles of the population S t n 5 ~ts/0. and is generally more conservative than a parametric procedure.3. may be approximated by using the results of some other test program. S D 3. s/ X¯. appropriate statistical tests shall be employed to substantiate this choice along with measures of test adequacy (2.). If samples are selected from inventory. it may be advisable to sample with unequal but known probabilities.1 Selection of a sample size depends upon the property or properties to be estimated. the principles in 3. a month.3 To determine sample size based on a tolerance limit (TL). (3) species or species group. However. a near-minimum property is generally the objective. For example. 3. Different confidence levels may be suitable for different products or specific end uses. 3.05 = precision of estimate. 6.3 Sequential Sampling—When trying to characterize how a certain population of lumber may perform in a structure. current design practices. as inferences made pertain only to that population. s and X¯. 3. see 3. etc. For the four other allowable stresses. This equation assumes the data is normally distributed and the mean is to be estimated to within 5 % with specified confidence: inaccurate results. random number tables may be used to determine which pieces will be taken for the sample.4. nonparametric or parametric techniques are applicable.3. X¯ = specimen mean value. 3. 4. and the sampling method shall be completely reported.2 with all members of the population sharing equal probability of selection.2.2 Population: 3. and t = value of the t statistic from Table 1. centers of distribution. Alternatively a nonparametric approach requires fewer assumptions. (2) geographical area over which sampling will take place (nation. Samples may be collected from stock at mills. The confidence with which this inference is to be made is a separate statement. When sampling current production.3. state. C. When this is not the case. NOTE 2—More details of this two-stage method are given in Ref (8). The principles of Practice E 105 shall be maintained. CV = coefficient of variation. and (6) moisture content. 3. The choice of a specified content and confidence is dependent upon the end-use of the material. is an estimate of the proportion of the population that lies above the tolerance limit. Note 1). and the precision with which the property is to be estimated. Where this is done. the sampling program should consider the location and type of log source from which the pieces originated. Such a composite sampling unit might consist of a sequential series of pieces chosen to permit estimation of the properties of the unit as well as the pieces. mill.2 apply to these composite sampling units and the sampling method shall be completely reported.1 and 3.05 CV D 2 (1) where: n = sample size. 0.167 t = 2 2 2 n = 5 44. the general principles of Practice E 105 shall be maintained. s = standard deviation of specimen values.2. 3.3. NOTE 3—The content.1 Random Sampling— The sampling unit is commonly the individual piece of lumber. etc. where the value of t is taken from Table 1. economic considerations. (4) time span for sampling (a day’s production.4. etc.3. if a parametric approach is used. the t statistic can be approximated by 2.622 ~45 pieces! 0. 3. the nonparametric tolerance limit concept of Ref (8) 2 .4. the actual variation in properties occurring in the population. it may be deemed more appropriate to choose a sampling unit. The sampling shall assure random selection of sampling units from the population described in 3.4 Sample Size: 3. the first sample was insufficient.1 It is imperative that the population to be evaluated be clearly defined.3. code requirements. If the second sample size exceeds the first. or CV.1 To determine the sample size for near-minimum properties. 7). 3. X¯. such as a package. Where this is done. a year.4. a tolerance limit with a content of 95 % describes a level at which 95 % of the population lies above the tolerance limit. Often the values of s. s = 300 000 psi (2067 MPa) X¯ = assigned E of the grade = 1 800 000 psi (12 402 MPa) CV = (300 000/1 800 000) = 0.05 X¯! 2 5 0. Therefore. that is more representative of how the lumber will be selected for use. (5) lumber size. it may be necessary to specify ( 1) grade name and description. NOTE 1—An example of initial sample size calculation is: Sampling a grade of lumber for modulus of elasticity (E). Commonly the mean modulus of elasticity and the mean compression perpendicular to the grain stress for the grade are estimated. Assuming a 95 % confidence level. In order to define the population.05 3 0.3.3 Sampling Procedure: 3.2 Determine sample size sufficient for estimating the mean by a two-stage method. obtain and test the additional specimens.3. a content of 95 % and a confidence level of 75 % may be appropriate for a specific property of structural lumber. For example.167 Calculate the sample mean and standard deviation and use them to estimate a new sample size from Eq 1. or they may simply be guessed (see example. at points of end use or directly from current production at the grading chains of manufacturing facilities. including types of processing methods or marketing practices with respect to any influence they may have on the representative nature of the sample. Appropriate content and confidence levels shall be selected before the sampling plan is designed. 3. refer to Practice E 105 for a recommended sampling procedure (see Appendix X3 of this practice for an example of this procedure). and t or CV and t are not known before the testing program begins. 5.

then a tolerance limit with stated content and confidence can be obtained for any sample size. For other tolerance limits or confidence levels.841 4.499 3.093 2.177 1. beginning with the smallest.869 for n = 30 Therefore n ' 30 specimens.074 2.160 2.145 2.055 3. desirable to select a sample size as large as possible commensurate with the cost of sampling and testing (see also 4.176 1.189 1.056 2.214 1. A CV of 22 % and a mean C11 of 4600 psi are assumed based on other tests. although 30 specimens is sufficient to estimate the 5 % PTL with 75 % confidence.178 2. and the confidence is greater than the nominal value. 27 and 28 for the first order statistic at 75 confidence).571 63.787 26 27 28 29 30 1.221 2. if normality is assumed.064 2. however.306 2. This will provide the sample size suitable for several options in subsequent near-minimum analyses.779 2.414 1.194 1.704 2. CV = 0.069 2.776 2.898 2. n.179 1.355 3. and the standard error (SE) of this statistic may be approximated by the following equation: NOTE 4—An example of sample size calculation where the purpose is to estimate a near minimum property is shown in the following calculation: Estimate the sample size. C The rank of the ordered observations. therefore. It is. with this size sample.22 X¯ = 4600 psi (31. SE 5 s Œ 1 K2 n 1 2~n 2 1! (2) where: s = standard deviation of specimen values.174 1.617 2.012 2. Although the frequency with which the tolerance limit will fall above (or below) the population value.925 5.187 1.175 1. The target PTL of the lumber grade is 2700 psi. 3 . may be chosen to make this quantity sufficiently small for the intended end use of the material (Note 4).209 1.254 1.179 2.021 2.182 2.162 1. (see Ref (8)).2 If a parametric approach is used.042 2.183 1. n = sample size. 3. corresponding to the required content. B Where the sample size falls between two order statistics (for example. That is.197 2.086 2.771 2.185 2.980 1.131 3.365 2. For calculating other confidence levels.182 1. The PTL is to be estimated with a content of 95 % (5 % PTL) and a confidence of 75 %.D 2915 – 03 TABLE 1 Values of the t Statistics Used in Calculating Confidence IntervalsA df n−1 A CI = 75 % CI = 95 % TABLE 2 Sample Size and Order Statistic for Estimating the 5 % Nonparametric Tolerance Limit.228 3.1 applies.3. the larger the sample size the more likely the tolerance limit will be close to the population value.423 1.4.878 2.4. n.576 95 % Confidence 99 % confidence Sample SizeB Order StatisticC Sample Size Order Statistic Sample Size Order Statistic 28 53 78 102 125 148 170 193 215 237 259 281 303 325 347 455 562 668 879 1089 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 20 25 30 40 50 59 93 124 153 181 208 234 260 286 311 336 361 386 410 434 554 671 786 1013 1237 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 20 25 30 40 50 90 130 165 198 229 259 288 316 344 371 398 425 451 478 504 631 755 877 1115 1349 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 20 25 30 40 50 A Adapted from Ref (12).797 2.052 2. see Ref (8).707 3. the PTL estimated by test may not be as close to the true population fifth percentile as desired.080 2.807 2. for a compression parallel strength test in which normality will be assumed.173 2.344 1.301 12. although the frequency that the tolerance limit falls above (or below) the population value. A larger n may be desirable.048 2.22) (4600) = 1012 psi (7.1 MPa! Consequently. the parametric tolerance limit (PTL) will be of the form PTL = X¯ − Ks.045 2.447 2.150 2.8772 1 30 2~30 2 1! (3) 5 310.250 3.303 3. the limitation expressed in 3.180 1.7 MPa) s = (0. the standard error (approximately 12 % of the PTL) illustrates that. the larger of the two is shown in the table.604 4.101 2.230 1.110 2.977 2.660 2.240 1. The sample size. Adapted from Ref (8).000 1. SE 5 1012 Œ 1 1.5 psi ~2.0 MPa) K = ( X¯ − PTL)/s = 1.032 6 7 8 9 10 1.877 From Table 3: K = 1.120 2.3.7).657 9.861 2.947 16 17 18 19 20 1.169 11 12 13 14 15 1.273 1.191 1.921 2.106 3.756 2.156 1.831 2. is controlled by the confidence level selected. NTLA 75 % Confidence CI = 99 % 1 2 3 4 5 2. the probability that the tolerance limit will be close to the population value depends on the sample size.167 1.960 2. see Ref (12) or (8). and K = confidence level factor. For example.262 2.891 2.763 2.706 4.200 1.204 1. may be used (Table 2).201 2.750 40 60 120 ` 1.060 2.604 1. corresponding to the required content is controlled.845 21 22 23 24 25 1.

985 1.413 5.462 2.244 3.845 0.380 1.376 1.25) 1−p 95 % Confidence (g = 0.909 0.783 0.733 0.805 2.138 1.732 0.997 7.114 2.768 0.336 6.002 1.769 0.415 3.821 2.521 2.431 2.568 2.423 2.354 1.712 0.960 1.946 0.042 2.282 1.324 1.681 2.763 1.407 1.129 2.514 1.103 4.336 1.151 1.340B 1.296 1.748 3.765 1.850 1.223 16 17 18 19 20 0.496 1.398 1.758 2.702 1.868 1.852 1.144 3.390 5.847 2.241 3.527 2.964 1.615 2.585 3.906 2.887 1.069 2.290 3.312 2.834 1.376 1.124 4.366 1.242 1.739 7.677 2.044 5.411 1.850 120 140 160 180 200 0.207 3.849 1.995 2.693 1.853 1.648 1.638 2.932 2.337 1.982 2.708 0.681 2.263 3.362 17.99 3 4 5 1.400B 2.454 2.400 2.518 2.019 0.332 1.085 6.729 4.425 0.664B 1.395 2.152 2.346 1.960 1.016 1.419B 2.811 2.391 0.094 2.777 3.669 1.686 1.932 0.895 1.729 1.556 1.723 0.617 2.430 2.476 1.424 2.973 3.145 4.121 1.824 2.565 1.387 2.722 0.657 2.893 3.952 1.757 0.251 2.963 2.742 0.957 1.657 5.712B 1.893 0.188 3.984 0.705 0.154 2.353B 2.700 2.816 1.817 0.694 0.299B 1.729 2.156 4.710 2.722 1.074 2.758 2.95 0.326 1.528 1.194 1.450 1.315 1.924 1.398 1.778 1.058 1.699 1.446 2.502 2.545 1.508 1.776 1.438 1.493 2.891 1.210 2.645 2.944 1.329 2.443 2.609 1.609 2.799 1.860 3.452 2.103 1.479 2.758 0.392B 2.732 1.2.354 1.727 3.794 0.203 10.853 3.727 1.450 1.726 4.941 6 7 8 9 10 1.697 1.411 3.449 1.405 2.083 1.269 3.350 1.292 3.683 1.854 0.234 3.336 2.883 0.961 3.758 0.724 2.580 23.553 500 600 700 800 900 0.894 0.879 0.912 5.607 1.794 0.974 2.908 0.857 1.033 2.974 1.916 1.809 1.542 1.901 1.570 0.783 0.358 1.95 0.180 2.924 2.385 1.863 1.645 2.326 0.927 2.869 1.660 3.714 1.634 4. 4 .524 2.371 1.837 0.708 0.063 4.991 2.766 2.775 0.396 3.736 0.775 0.365 2.911 0.978 0.416 1.310 2.595 2.393 1.342 1.335 3.087 1.926 2.328 1.430 1.048 2.398 1.723 2.738 1.643 4.646 2.037 3.104 3.841 1.753 1.898 2.125 1.849 0.382 2.464 4.777 2.431B 2.326 n A B Obtained from a noncentral t inverse approach.401 1.830 0.764 1.220 1.762 0.431 1.475 2.75 0.539 1.742 8.954 4.412 1.255 1.153 2.577 1.447 2.930 0.733 0.763 1.577 2.994 0.714 3.431 2.282 1.964 1.90 0.95 0.276 3.671 2.703 0.142 2.514 1.411 2.75 0.189 2.849 1.674 1.314 2.919 0.737 2.965 1.745 0.588 2.371 3.568 2.353 1.869 1.707 2.536 1.90 0.483 1000 1500 2000 2500 3000 inf 0.555 7.710 1.357 2.801 1.364 1.464 1.320 1.249 3.745 0.741 1.007 2.026 2.674 1.826 0.825 0.434 1.732 2.790 1.586 2.736 1.708 1.756 2.728 1.802 0.711 0.767 2.391 1.941 2.908 1.740 0.406 1.722 0.302 1.386B 2.671 2.524 2.862 1.778 1.417 1.154 1.832 21 22 23 24 25 0.695 2.732 1.560 1.039 2.286 3.385 1.99 0.688 0.374 2.739 2.501 2.768 0.149 6.355 3.406 2.331 3.770 1.689 0.340 1.253 2.990 1.05) 99 % Confidence (g = 0.048 5.940 1.753 1.300B 1.506 2.591 1.465 3.064 2.593 2.522 3.557 3.220 2.868 1.614 2.999 0.813 5.407 4.361 2.797 2.708 3.497 1.372 2.351B 2.247 1.167 2.645 2.703 1.869 1.715 0.691 0.701 0.445 1.381 5.162 3.697B 1.057 3.065 3.507 2.679 250 300 350 400 450 0.852 2.895 2.456 2.010 0.75 0.365 1.617 1.371 0.461 1.438 2.495 2.923 2.332 1.458 1.391 1.838 2.978 2.900 1.788 1.113 1.189 3.876 0.726 3.407 7.823 2.770 2.682 3.355 4.043 1.374 9.023 1.008 1.809 0.90 0.634 3.744 1.356B 2.812 0.818 0.522 2.869 0.369 1.363 1.025 0.692B 1.864 1.902 1.787 1.740 0.788 1.369 2.698 0.D 2915 – 03 TABLE 3 K Factors for One-Sided Tolerance Limits for Normal DistributionsA 75 % Confidence (g = 0.539 1.99 0.182 1.602 30 35 40 45 50 0.689 1.663 2.498 3.397 3.552 1.795 1.490 2.719 0.783 1.329 1.241 2.859 1.836 1.426 1.739 1.203 2.400 3.532 1.808 2.703 1.684 1.200 1.763 1.355 1.453 13.640 3.696 2.689 1.134 1.666 2.674 1.447 3.181 3.669 1.672 1.460 1.726 2.521 1.747 1.311 1.151 2.712 1.697 1.01) 0.646 1.313 1.949 1.582 2.542 2.032 2.389 8.602 2.166 2.030 1.126 2.412 3.182 3.326 0.650 2.484 0.436 0.851 0.662 1.858 0.815 1.879 1.075 11 12 13 14 15 0.378 2.386 1.471 1.675 2.773 1.282 1.769 2.453 2.470 0.977 1.702 1. Computed using formula X5.942 1.276 2.581 1.623 2.800 1.028 2.958 1.756 2.422 3.552 2.349 2.830 4.093 2.644 1.762 1.521 1.666B 1.502 1.326 1.296 1.051 1.927 1.686 2.306 1.679 1.585 2.794 0.890 0.476 2.854 2.846 0.209 2.884 2.083 2.727 0.465 3.159 1.109 2.881 0.740 1.555 1.639 2.503 1.411B 2.808 4.797 2.625 1.838 2.822 1.125 60 70 80 90 100 0.475 1.566 3.816 2. see Ref (15).962 3.487 2.727 1.750 0.516 2.715 0.358 1.485 1.317 1.267 2.604 1.414 2.541 2.127 3.906 1.886 2.323B 1.974 0.480 1.898 2.733 2.899 1.156 2.460 2.719B 1.777 1.338 4.932 0.489 2.900 12.319 2.473 4.699 1.777 0.733 1.

it is necessary to account for the effect 6 A uniform load distribution is commonly encountered in use. and Ki = values are given in Table 4. the procedures shall be fully described. For this reason. To determine the apparent modulus of elasticity. Also. These apparent moduli may be standardized for a specific span-depth ratio and load configuration. 3.20 0. NOTE 5—A best-fit procedure should recognize the low power of some published procedures. such as knots.960 . 4.3. Factors to adjust Eai for span-depth ratio and load configuration may be derived from Eq 4. When this is the case. only one sample size need be used. Experimental evidence suggests that these equations produce reasonable results with solid wood when converting between load conditions at a fixed span-depth ratio. when the modulus. to adjust Eai to standardized conditions.1 Often. To check the fit. a description of the selection procedures and a tabulation of distribution parameters shall be provided. When adjusting shear strength and compressive strength perpendicular to the grain the procedures of Practice D 245 shall be used. Limited data indicate the ratio of E/G for individual pieces of lumber can vary significantly from this value depending upon the number. separate justification shall be provided for factors used to adjust test values to standardized conditions. the procedures given in Practice D 1990 shall be used for dimensions.2 and 3. the series of tests outlined in Ref (10) represents several alternatives. as in the sampling of some or all mills in a defined region. size. 7). and the spans over which deflections are measured (13). historically it has been assumed that the modulus of rigidity (G) is one sixteenth the shear free modulus of elasticity ( E). if the span-to-depth ratio is the same. and exponential distributions (5). G = modulus of rigidity. solve the equation: 3. tensile strength. these adjustments decrease in accuracy with increasing change in moisture content. S DS D S DS D h1 1 1 K1 L 1 Eai2 5 h2 1 1 K2 L 2 2 2 E G E G Eai (4) where: h = depth of the beam. The apparent modulus.3 In calculations using Eq 4 and that involve mean trends of large populations.3. which customarily serve as a basis for design range. A 5 Deflection measured at midspan midspan load points midspan load points midspan See Appendix X4 for an example of use of Table 4. and location of knots present. If parametric procedures are to be used. (Ref (11)). 7 When using these conversion equations with solid wood. the slope of grain in the piece. revised standard tables of the KolmogorovSmirnov statistic are presently available only for the normal. 4. For instance. therefore. 4. and compressive strength parallel to the grain. E ai.3. may be affected by the location of growth characteristics. with respect to loads and supports. The equations were derived using simple beam theory for a simply supported beam composed of isotropic. Analysis and Presentation of Results 4. special procedures may be required. bending strength. based on some other set of conditions is known.4.080 0.939 1. In the case of multisource data. but may be closely approximated by applying the load at the one-third points of the span. 4. a single E/G ratio is usually assumed.6 When tests at standardized conditions of load and span are not possible. Any “best-fit” judgment (Note 5) between competing distributions shall be documented. and (2) an appendix of unadjusted individual test specimen results. however. those based on the methodology introduced in Ref (9).4 Often the objective of the evaluation program will be to estimate mean and near-minimum properties simultaneously. conditions of anticipated end use. It should be noted.D 2915 – 03 of shear deflection on beam deflection. homogeneous material. TABLE 4 K Factors for Adjusting Apparent Modulus of Elasticity of Simply Supported BeamsA Loading Concentrated at midspan Concentrated at third points Concentrated at third points Concentrated at outer quarter-points Concentrated at outer quarter-points Uniformly distributed 5 Spans.3. then the required sample size must be determined by procedures that are statistically appropriate for the sampling method chosen. logistic. This load configuration is difficult to apply in testing.5. modulus of elasticity. based on any set of conditions of span-depth ratio and load configuration. L = total beam span between supports. as far as possible.873 1. 4) have been shown to be among the more powerful tests (6.7 If this assumption is critical for the intended 4. used in deflection equations that attribute all deflection to moment.5 If a sampling unit other than an individual piece of lumber is to be used. and that adjustments for more than five percentage points of moisture content are to be avoided.4. for example. tests based on the AndersonDarling statistic (2.4. E = shear free modulus of elasticity. lumber is not homogeneous within a piece with respect to modulus of elasticity.2 Properties shall be adjusted to a single moisture content appropriate for the objective of the testing program. Although test results can be adjusted for moisture content. Eai2.3 Modulus of elasticity values of primary concern are apparent values.3. 4. When adjustments are required.1 The results of the tests performed in accordance with Test Methods D 198 or other standard testing procedures shall be analyzed and presented as (1) a set of summarizing statistics. as provided for in 3. Ki 1.200 0. Standardization should reflect. It is further cautioned that conversions may be less appropriate when converting between edgewise and flatwise specimen orientations. 3. Eai. Care must be exercised when converting between different span-depth ratios to assure that the adjustments are appropriate for the end use. it is suggested that the specimens be conditioned as closely as possible to the target moisture content prior to test. It should be the greater of the two obtained in accordance with 3.2 If modulus of elasticity results are not measured at standardized conditions. go from 17 to 21 times the depth of the specimen.4. that not all tests are valid for all distributions and that these procedures are effective for checking central tendency. In all cases.3.

5. or an empirical cumulative distribution function. Adequate significant digits shall be maintained in all intermediate calculations to avoid rounding errors in the statistics. shall be presented. the first order statistic is the lowest test value or the weakest piece in the sample. a given percent of all intervals found in this manner are expected to contain the true population mean. and is given in Table 1. if the population is normally distributed. If parametric procedures are used for analysis. x2. or combination thereof. For example. arrange the test values in ascending order.1-4. 4.8. That is.5. maximum widths are given in Table 6. confidence intervals. 4.2 The sample standard deviation is calculated as follows: ( ~xi 2 2 @~( xi! 2/n# i51 @ x j 2 x ~ j 2 1 !# 1 x ~ j 2 1 ! where k is the desired percentile point estimate sought. this is done using the results of 4. 4.5. 4. 4. where: xi = individual observations.5. They are taken from Practice D 245. for each successively higher value until i/(n + 1) $ k/100. which includes a safety factor and a 10-year cumulative duration of load effect (normal loading).5. Beginning with the lowest value (that is.1 1/1. 4.6. 12). the second order statistic is the second weakest piece.5.5. 4. call it the jth value. Symbolically.. Interpolate the nonparametric k percentage point estimate by: TABLE 5 Reduction Factors to Relate Test Statistics to Allowable Properties Property Modulus of elasticity Bending strength Tensile strength Compressive strength parallel to grain Shear strength Compressive strength perpendicular to grain (8) NOTE 6—Order statistics are ranked test values from the lowest to the highest. 8.5.5. the corresponding NTLs can be determined (8. 4.3 The confidence interval (CI) for the mean is calculated as follows: CI 5 X¯ 6 ~ts/= n! 4. if the sample size was 93 and the confidence level was chosen to be 95 %. then the calculation of sample means.5.5. A CI of this type provides that. standard deviations. first order statistic.D 2915 – 03 F k NPE 5 100 ~n 1 1! 2 ~j 2 1! application.4 The sample nonparametric percent point estimate (NPE) may be interpolated from the sample. equals or exceeds the sample k percentile point estimate.5.5 Statistics shall be shown with three significant digits. 4.7 A histogram.34) .5.3. The accuracy of existing near-minimum properties may be assessed using the results of 4.4) 50 (0.1 1/2. m = 2. The sample mean is an unbiased estimator of the true population mean.1 The sample mean is calculated as follows: 4. If other lower percentiles are estimated. . Table 2 depicts the order statistic required to determine the lower-5 % NTL at a given sample size and three confidence levels.5. x3. psi (MPa) 100 000 (690) 500 (3. the testing program bears out the value allowed for the population with the associated confidence.3.5. 4. the testing program may bear out the existing values.5.9 1/2. the lower-5 % NTL with at least 95 % confidence would be the second order statistic.6 If parametric methods are used. 12). tolerance limits. where i is the order of the value.4) 500 (3. the parametric point estimate (PPE) and lower parametric tolerance limit (PTL) shall be estimated by procedures documented as adequate for the method adopted (1.6. but no confidence statement may be (7) where t depends on the sample size and confidence level.4) 500 (3. or both..67 Modulus of elasticity Bending strength Tensile strength Compressive strength parallel to grain Shear strength Compressive strength perpendicular to grain 6 Class Width.4 The adjustment factors used to reduce the test statistics to the level of allowable properties depend on the property and are shown in Table 5.5 are given in Appendix X1 and Appendix X2.8 If a sampling unit other than an individual piece of lumber is used.4.4. and exclusion limits must be made in a manner statistically consistent with the sampling procedure chosen. and n = sample size. If an allowable mean property for a population falls within the confidence interval obtained in accordance with 4.1 1/1. call them x1.34) 50 (0. either a cumulative distribution function or a probability density function can be shown superimposed on the empirical cumulative distribution function or the histogram respectively. see Note 6). where m depends upon the sample size and confidence level.5. (5) Œ Œ n s5 5 n–1 or (6) n ( ~xi – X¯!2 i51 n–1 NOTE 7—Two examples of typical test data and a summary of the results that meet the requirements of 4. xn. If the existing property falls at or below the point estimate as calculated in 4.8. n ( xi/n i51 X¯ 5 G TABLE 6 Maximum Class Width to Be Used in Histogram Plots Property Factor 1 1/2. To perform the interpolation.5 The nonparametric lower tolerance limit (NTL) of a specified content is the mth order statistic. it is recommended that the moduli of elasticity and rigidity of the individual pieces be measured (see Methods D 198). or 4.5.5. The class widths for a histogram depend on the property. calculate i/(n + 1). 4. etc.6 If the purpose of the testing program is to evaluate the accuracy of existing allowable properties for the population sampled.5. For example. 4. and 4. or 4.

1. X1. it was decided this sample E did not verify the design E. testing.5.1 lumber.4 The end use of a specific product will dictate the specification requirement. if (NPE − NTL)/NPE < d.3 Results obtained following the procedures and analyses of this practice may also be used to characterize the population sampled for establishing design values. and 4. one should be cognizant of the imprecision in the NTL consequent on the sample size (see 3.5. Analysis of the tension strength values was conducted in accordance with 4.2) did not contain the value as printed in Table X1. 4. structural lumber. predetermined by the user will normally be in the range from 0. and Weibull distributions of edgewise bending E and tensile strength are shown in Fig.01 to 0. Indeed this practice addresses itself to the procedures for sampling specified populations and procedures for interpreting the results.5.7 If the purpose of the testing program is to establish allowable properties for the population.4 for tensile strength can be estimated directly from Table X1.3.2. only 7 . or the NTL may be used for the allowable value.5. if the relative difference between the NPE and the NTL is sufficiently small. the content.1.3). Generally. knot location.6. is not shown in Table X1.4). from one mill. only the best fit distribution need be illustrated. and the costs of sampling.5.D 2915 – 03 associated with this conclusion. The representativeness and size of the sample influence how the characterization can be made. F t. 4. The 80 test specimens were equilibrated to an average of 15 % moisture content (see Note 4 and Practice D 245).5. If this condition is not satisfied.2 The purpose of the test was to evaluate the bending modulus of elasticity.3 The design value for the grade and species sampled is given in Table X1. additional samples must be taken as described in 3. Provisions are made for estimating both the mean and near-minimum property values. of a one-mill sample relative to present design values.5.1. The allowable value of modulus of elasticity shall be the sample mean of 4.3. and confidence associated with the test sample. X1. the PPE and PTL of the parametric procedures provided for in 4. If the latter course is chosen. Applications 5. Note that the nonparametric estimates in Table X1.1 and Fig. such as the mean or a near-minimum property. If this condition is not satisfied.10). other interpretations may be appropriate.4. Contemporary practice is reflected in 4.7. 2 grade Hem-Fir two-by-fours (current lumber agency grade rules). the confidence interval for the mean E value (Table X1. It may be desirable to tabulate additional information.1.7 Using 4. or incorrect decisions. reduced to the level of allowable properties.4. X1.5. 6.5. most of which. and analysis.1). and tensile strength. depending on purpose. X1. 1152/ X1.1. the risks that surround imprecise estimates.1 Population Description—Selected at random.5 Several of the individual test results are shown.1 The results.10). if ts/( X¯ = n ) # l . X1. 4. It cannot be implemented without the selection of values for the confidence levels and degree of precision needed at various stages of the procedures.3. wood APPENDIXES (Nonmandatory Information) X1.3. Alternatively.5.8. Keywords 6.5. such as specific gravity. 4. Note that tensile strength data is ordered in ascending order.5. however. (that is..3. Consequently the fifth percentile estimate will be considered for strength and the mean value for E (see 4. Consequently.4 and 4. this is done using the results of 4. X1. 5. but the estimates for modulus of elasticity are based on data. 5. 4.5.6 may be employed in a parallel manner. or 4.01 to 0.5. The specific characterization with respect to the population.5. where d will normally be in the range from 0.3 until the condition holds. 4. 4. 5.2 Where properties have been previously assigned to a lumber population.6. may be used to evaluate the accuracy of existing allowable properties or to establish allowable properties. additional samples may be taken until the condition holds.4. TYPICAL EXAMPLE—COMMODITY LUMBER as an example of data that is typically recorded (Table X1. one purpose of this practice is to provide a format for evaluation of this assignment through full-size lumber tests.2. A table of test statistics is given in Table X1. or 4.6. however.7). etc.5. the existing value must fall below the tolerance limit as calculated in 4.6 If appropriate best fit tests have been carried out and documented. After adjusting the nonparametric lower -5 % tolerance limit to an allowable design value (that is.5. 4.8. These values should be given careful consideration so that they are compatible with the anticipated end use. 5.5. were 80 No. lognormal.5. the allowable value of any nearminimum allowable stress shall be the sample 5 % NPE of Section 4. depends on the objective. where l. This condition is essentially that of having sufficiently narrow confidence interval for the NPE. X1.2. if the width of the confidence interval is a sufficiently small fraction of the mean (for example. In order to associate a confidence statement.4 Histograms and fitted normal. illustration of other options is instructive (see Table X1. E. X1.

psi (MPa) 675 (4.4.2 Tensile Strength (1000 psi) TABLE X1. it can be seen that this value is below the value shown in Table X1. Species/Grade Hem-Fir No. 2 Grade Hem-Fir Two-byFoursA 2.1 (Ft = 675 psi (4. 8 Ft. therefore. the sample tension values do not verify the design tension value.1 = 548.D 2915 – 03 FIG. . 2 A X1.1 Design Values for No.6) Design Values E.1 Static Edgewise Modulus of Elasticity (106 psi) FIG.6 psi (3.8 MPa)). X1. X1.8 Similar analyses could be performed using parametric procedures and employing the values shown in Table X1.6 MPa)). psi (MPa) 1 400 000 (9 646) National Design Specification for Wood Construction.

47 (88) 3. C Adjusted to .8) Sample Size 80 80 A All statistics in psi. psi (MPa) Tensile StrengthA. Histograms of test results are shown in Fig.2 %. in./d of 21 and a uniform load.2 Example Test Results for No.2. Obtained by 5. quarter-point load.3 contains estimates of near-minimum values.D 2915 – 03 TABLE X1.0 15. Actual average moisture content of specimens equaled 11. (mm)B Thickness.0) 1.8340 (5746) 1.G.1 Population Description—(Species) ladder rail stock graded in accordance with the (Grading Rules) as “V. %C 15. B 95 % confidence.0) 1257 (8. psi (MPa)B 1 713 200 (11 804)–1 797 400 (12 384) 9 758 (67) 9 520 (66)–10 014 (69) Mean.47 (37) 1.230 (8. X2.9) 1169 (8.2.2.46 (37) 1. C Adjusted to .51 (38) 1.42 (87) 3. (mm)B Bending Strength Ratio. Only mean and lower tail properties estimated by nonparametric procedures were of interest.8091 (5575) 0.580 (4. D Adjusted to 12 % moisture content./d of 21 and uniform load.47 (88) 3. by 8 ft were selected randomly from stock at a ladder manufacturer in (location).726 (5.0 13. Examples of individual specimen data are shown in Table X2. At test moisture content.8745 (6025) 0.6) Sample Size 200 200 .152 (7.6) Confidence Interval for Mean. B TABLE X1.5) 0.45 (88) 3.4 Estimates of Population Parameters for Two-by-Four Sample Parameter Weibull: 5 % point estimate Lognormal: 5 % point estimate 5 % TL (75 %) Normal: Mean Standard deviation 5 % point estimate 5 % TL (75 %) Nonparametric: 5 % point estimate 5 % TL (75 %) A Static Edgewise Modulus Elasticity 106.7) 994 (6849) 959 (6607) 1061 (7310) 667 (4596) 950 (6545) 3. testing.4 %. psi (MPa) 0.208 (8.6)–1 350(9.0) 1.8255 (5688) 1.7790 (5367) 2.3 Example of Test Results Ordered by Tensile Strength—Two-by-Four Sample Specimen Number 1 1 1 1 1 A B C Moisture Content at Test.5) 1152 (7. TYPICAL EXAMPLE—LADDER RAIL STOCK properties.46 (88) 1. Table X2. reductions in accordance with 4.616 (18.8549 (5890) 0.3) 1. Empirical cumulative distribution functions are TABLE X2.4. The purpose of the sampling.9) 1092 (7.3) Standard Deviation. and analysis was to obtain the bending modulus of rupture (MOR) and modulus of elasticity (E) of typical ladder rails for use in a research study on ladder rail X2. psi (MPa) 301 500 (2077) 1 836 (12. psi (MPa) 1 201 600(8 279) 1 250(8.49 (38) 13 13 52 47 52 P 43 P1 P 15 P 28 P 22 Test . 9 Standard Deviation. Specimens were equilibrated in a conditioning room.1 Ladder Rail Test StatisticsA Property Static edgewise modulus of elasticityC Modulus of ruptureD 1 755 300 (12 094) Confidence Interval for Mean.169 (8.0) 0. psi (MPa)A Width./d of 21. Ladder Rails.2 Data reduced to summary statistics are shown in Table X2.8490 (5850) 1. 2 Grade Hem-Fir Two-by FoursA Property Static edgewise modulus of elasticityC Tensile strength Mean.7) 1.3.0) 0. corrected to . uniform load. psi (MPa) Edgewise Modulus of Elasticity 102.149 (7.1 and Fig./d of 44. psi (MPa) A All statistics in psi.1.0 1004 (6. 95 % confidence. The 95 % confidence level was deemed appropriate for both E and MOR in this study. X2.9) Not reduced to allowable property. The standard deviation was 1.1 of Practice D 245. and 12 % moisture content.9) 0. in. X2.50 (38) 1.270 (8. TABLE X1. % Tensile Strength.” Two hundred pieces of 13⁄8 by 23⁄4-in.0 15.4 (not rounded). all adjusted to 15 % moisture content in accordance with 4.2016 (8279) 0. X2. psi (MPa)B 1 148 500(7 113)–1 254 700(8 645) 1 100(7. psi (MPa) 238 500(1 643) 547(3.2385 (1643) 0.0 16.

7 for the MOR.8 11. Consequently.6 14 343 (99) 11 423 (79) 6 505 (45) 9 708 (69) 2.3 and Fig. comparisons between the NPE and several NTL’s can be made (Table X2.7 was chosen). Therefore.2.17 > 0. % Modulus of Rupture. Either more sampling (see 4.29 (8888) 1. B psi (MPa).406 (36) 1.7) is required or the NTL (5364 psi (37 MPa)) may be used as the best estimate of the population MOR.23 (8475) 1.4./d of 21 and uniform load. actual conditions were . ts/ X¯ = n = 0. (mm) 103 111 114 121 12. adjusted to .3). X2.1 Edgewise E (10 6 FIG. the research suggested an edgewise E of 1.7 3 106 psi could be used as a design value (Practice D 245 rounding rule would round the test value to 1.B Width at Test. in.760 (70) 2./d of 33 and center point load. adjusted to 12 % moisture content in accordance with 4.366 (35) 1. not adjusted to allowable properties.30 (8957) 1. not reduced to allowable property.0 8. not reduced to allowable property.753 (70) 2.381 (35) 1.762 (70) 1.05) with 95 % confidence.8 but this would be out of the confidence interval for the mean.16 (7992) 6518 (45) 6072 (42) 5364 (37) 5353 (37) 10 psi (MPa).2 Bending Strength MOR (1000 psi) psi) shown in Fig.D 2915 – 03 TABLE X2.2.5. Adjusted to . X2.37 (9 439) 2. X2. in./d of 21 and uniform load in accordance with 4.10) the relative difference for the NTL at a 95 % confidence level does not meet the criterion (65185364/6518 = 0.7 it was determined that the dispersion of E (static edgewise) measurements met the 5 % requirement (that is. A 6 FIG.3 Following the procedures of 4.386 (35) Statistics adjusted to 12 % moisture content in accordance with 4.4 Continuing the procedures of 4.6 11.10). thus 1. psi (MPa)A Edgewise Static E.17 (14 951) 2. the 95 % confidence level goal of X2.2.2 Sample Test Results—Ladder Rail A B Specimen Moisture Content at Test.1.80 (12 402) 1.1 for MOR is not met. TABLE X2. 106psi (MPa) A.51 (17 294) 1. X2. Maintaining the 10 % relative difference criterion (NPE-NTL/NPE < 0.024 # 0. 10 . X2. adjusted to 12 % in accordance with 4. X2.3 Estimates of Near-Minimum Population Parameters of Ladder Rail Property Edgewise modulus of elasticityA Modulus of ruptureB 5 % Point Estimates 5 % Tolerance Limits 75 % Confidence 5 % Tolerance Limits 95 % Confidence 5 % Tolerance Limits 99 % Confidence 1.784 (71) 2. (mm) Thickness at Test.

These are the ten random start points. 11 . from the green chain) at a manufacturing facility. 36. 10k = 50. 3. Select ten random numbers between 1 and 50. 40. X2. 47.1).1 When sampling from current production (that is. 29. 42.3 Empirical Cumulative Distribution Function for E FIG. Systemically select test specimens using an interval length of 10k beginning at each of the random start points (that is. 14. therefore. Let k = 5.4 Empirical Cumulative Distribution Function for R X3.2 Following the procedure outlined in Practice E 105 (A1. and 50 (Table X3. 31. the following procedure allows the estimation of a standard error (SE) of the estimate as well as some information about the within-andbetween sample variance. X2.D 2915 – 03 FIG.6) k is generally chosen to be five or greater. X3. EXAMPLE—SAMPLING PROCEDURE X3. 9. random start x + 10k).

36 86 136 186 . (9) Woodruff. .” Journal of the American Statistical Association..” Journal of the American Statistical Association.60 million psi. E2 5 ~1.. This process is continued until the desired sample size is obtained. K depends upon sample size n. “Estimation of Tolerance Intervals and Selection Procedures for Property Characterization.... Vol 47. A 9 59 109 159 . pp. H. 1952.” National Bureau of Standards Handbook . R. .. and Romig. A.D 2915 – 03 TABLE X3. . pp.S. . the formula can seriously overestimate the K factors. .. (6) Stephens. .. “Distribution of the Anderson-Darling Statistic. 193–212.60 million psi (X4. ONE-SIDED TOLERANCE LIMITS FOR A NORMAL DISTRIBUTION g = (4n − 5)/(4n − 4).. 583–588...939 Therefore. Forest Products Laboratory. From Table 4: h1/L1 5 1:14 E 1 5 1.. Vol 64. where: REFERENCES (1) Dodge. 1974. C. A. 591–595. Vol 69... U.. . 14 64 114 164 . D... Interim Report. “Goodness of Fit for the Extreme Value Distribu- tion. as well as percentile 100-p and confidence 1-g. .1 A one-sided tolerance limit. P. 1961. R.. .“ Confidence Intervals for Medians and Other Position Measures. 1954. The formula is as follows: PTL 5 ¯ X – Ks Z 5 T – ~b0 1 b1T 1 b2T2!/~1 1 b3T 1 b 4T2 1 b5T3! where: T = =Ln~1/Q2! (Q = p for Zp and Q = g for Zg) b0 = 2.. The MOE value obtained was 1.. S. is a value about which it may be said with confidence 1-g. G.” Biometrika. M.1) X5... Vol 23. .. W.. A... T. respectively. John Wiley and Sons. Sampling Inspection Tables. “Tests of Fit for the Logistic Distribution Based on the Empirical Distribution Function.” United States Department of Agriculture...432788 b4 = 0. Second Edition. 42 92 142 192 .. Government Printing Office. A.. A. M.. 1952. J.. .2) NOTE X5. NY. A. .. (10) Johnson. 47 97 147 197 . G..2 are approximations (see Ref (15)). T. H. (5) Stephens. . what would be the apparent MOE for loads applied at the one-third points of the span with a span-depth ratio of 21:1? Deflections were measured at the center of the span. WI. 1977..... pp. . 40 90 140 190 . A... Forest Service. For small values. 635–646. “A Test of Goodness of Fit.3) Zpg 1 =Zp 2g 2 2 @g 2 2 Z g 2/~2~n 2 1!!#~Zp 2 2 Zg 2/n! g 2 2 Zg 2/~2~n 2 1!! (X5. 1978. pp. 730–737. W..189269 b5 = 0.. M.. (4) Lewis. H..09796/1. pp. 1959.. “EDF Statistics for Goodness of Fit and Some Comparisons.” Annuals of the Mathematical Statistics. . (7) Stephens. calculated from the sample data.. . 1-p.1) where X¯ and s are the mean and the standard deviation... Vol 49....” Biometrika.802853 b2 = 0. K values are given in Table 3 or they may be calculated from the following formula: K5 (X5.” Journal of the American Statistical Association. pp. New York.1 Test Specimens to Be SelectedA 3 53 103 153 .” Annuals of the Mathematical Statistics. 1963. D.515517 b1 = 0. Vol 32.010328 b3 = 1. “Asymptotic Theory of Certain Goodness-of-Fit Criteria Based on Stochastic Processes. . F.. X4... “Graphical Solution for Shear Deflection of 12 . M.. (2) Anderson.. 29 79 129 179 .. (3) Anderson.. and Darling.. . W. Assuming an E /G ratio of 16:1... and Zp and Zg are calculated with the following formula: X5.60 million psi E/G 5 16 K1 5 1. “Experimental Statistics. 1979.001308 (X5.1 An average apparent modulus of elasticity was obtained by testing simply supported beams loaded at the center and having a span-depth (L /h) ratio of 14:1. 50 100 150 200 . ..20 K2 5 0.70 million psi (X4. Madison. and Haskell.2) E2 5 1. (8) Natrella... 765–769. of the population is greater than PTL. 31 81 131 181 ........ (11) Hilbrand.. and Darling. pp. that at least a proportion. PTL.. 118–121.034070!*1. . EXAMPLE OF USE OF TABLE 5 TO ADJUST MODULUS OF ELASTICITY (MOE) TO STANDARD CONDITIONS h2/L2 5 1:21 X4..1—K values computed using Eq X5. 91.. Vol 66.

Your comments will receive careful consideration at a meeting of the responsible technical committee. which you may attend. Hafner Publishing Co.” Report to ASTM task group investigating the validity of Table 2 of ASTM D2915 – 74. C. West Conshohocken. ASTM International takes no position respecting the validity of any patent rights asserted in connection with any item mentioned in this standard. Department of Agriculture. Irwin. L. Your comments are invited either for revision of this standard or for additional standards and should be addressed to ASTM International Headquarters. Forintek Canada Corp. (12) Guttman.S. D. 1967. United States. PA 19428-2959. Statistical Tolerance Regions: Classical and Bayesian. pp. J.org).org (e-mail). 1970. Darien. pp. If you feel that your comments have not received a fair hearing you should make your views known to the ASTM Committee on Standards. Forest Products Laboratory. 88–93. U. 61–62. I.. Madison. and the risk of infringement of such rights. either reapproved or withdrawn. Statistical Tolerance Regions: Classical and Bayesian. “An Equation for One-Sided Tolerance Limits for Normal Distributions. Darien. Carol L.. Wisconsin. Users of this standard are expressly advised that determination of the validity of any such patent rights. Forest Service. CT. 610-832-9555 (fax). Hafner Publishing Co.D 2915 – 03 Flexural Members.” Research Paper FPL 458.astm. or service@astm. or through the ASTM website (www.. and Barrett. PO Box C700. 13 . (14) Link. CT. (15) Guttman. 1985. (13) Palka. This standard is subject to revision at any time by the responsible technical committee and must be reviewed every five years and if not revised. 1985. at the address shown below. 100 Barr Harbor Drive.. 1970. “An Examination of E/G Values for Canadian Spruce Lumber. Vol 17. Individual reprints (single or multiple copies) of this standard may be obtained by contacting ASTM at the above address or at 610-832-9585 (phone). Report on file at ASTM Headquarters.” Forest Products Journal. are entirely their own responsibility. This standard is copyrighted by ASTM International. No. 6.