Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
10 CHAPTER 3- INTERLABORATORY TEST RESULTS AND ANALYSIS 3.1 Test Data The test data collected from laboratories included three replicate measurements of organic content of 12 soil-organic material blends consisted of three soil types and four percentages of ground walnut shells. This totaled in 36 measurement data from each laboratory. Among the 30 laboratories selected, 27 laboratories submitted full sets of measurements in every soil type. The measured data and the related statistics corresponding to clay, silt, and sand blends are provided in Appendices B, C, and D, respectively. Tables B-1, C-1, and D-1 of the appendices present the data and Figures B-1, C-1, and D-1 demonstrate the median, the maximum, and the minimum organic content values. To be able to compare the percentages of organic material measured with the percentages of organic material added, the measured organic content data were adjusted for the amount of organic material inherent to each soil. Table s B-2, C-2, and D-2 and Figures B-3, C-3, and D-3 provide the adjusted percent organic content data. In each of the figures, the middle point of each data point represents the adjusted median and the lower and upper bars represent the adjusted minimum and maximum data values, respectively. 3.2 Method of Analysis Test results of the ILS were analyzed for precision in accordance to ASTM E 691[2]. Prior to the analysis, any outlier data was eliminated by following the procedures described in E691 for determining repeatability (Sr) and reproducibility (SR) estimates of precision. For each set of data, the h and k statistics, representing the between and within laboratory consistency, were used to identify the outlier data. Data exceeding the critical h and k values were eliminated as described in Sections 3.3. Once identified for elimination, the same data were eliminated from any smaller subsets analyzed. Figures B-2, C-2, and D-2 of Appendices B, C, and D provide the graphical representations of the computed and critical h- and k- statistics. 3.3 Analysis of Results Multiple sets of data in each soil type were eliminated based on the critical h and k values. After eliminating the outlier data, the averages, the repeatability and reproducibility standard deviations of the data were determined. The eliminated unadjusted data are shown shaded in Table B-1, C-1, and D-1 and demonstrated in Figures B-2, C-2, and D-2 of Appendices B, C, and D. The eliminated adjusted data are shown shaded in Table B-2, C-2, and D-2 and demonstrated in Figure B-3, C-3, and D-3 of the appendices. The Sr and SR precision estimates were determined using the remaining data conforming to E 691 method. A summary of statistics of the measurements is shown in Error! Reference source not found.. The comparison of the design and measured organic content values in the table indicates that every soil has a certain percentage of inherent organic material. The inherent organic content of the soils are observed in the 0% column in Tables B-1, C-1, and D-1 of Appendices B, C, and D. As indicated, clay has the most amount of inherent organic material, whereas sand has the least amount. Upon subtracting the inherent organic contents from the measured organic
11 contents, the average of the measured values agree closely with the design values as shown in table of the adjusted values (Table 3-2). In addition to the adjusted averages, Table 3-2 also provides the adjusted variability of the measurements. It is indicated from Table 3-2 that the standard deviation of the measurements for sand increases with the increase in the percentage of organic material. The increased variability of the sand blend with higher percentages of organic material indicates segregation of organic material during shipment. This could be explained by the non-cohesive nature of sand that does not allow adhering sand particles to ground walnut shell grits. The tests of statistical significance in the next section would determine if the differences in averages and variability of measurements were significant for the different organic contents and different soil types. Table 3-1: Summary of Statistics of organic content measurements after elimination of outlier data Soil Type Design Organic Content No. of Labs Average Measured Organic Content Sx Repeatability Reproducibility 1s (Sr) D2s 1s (SR) D2s Clay 0% 27 3.03 0.981 0.277 0.785 1.018 2.880 Clay 2% 25 5.38 0.925 0.287 0.813 0.966 2.735 Clay 5% 26 8.29 0.985 0.259 0.732 1.017 2.879 Clay 8% 24 11.16 0.787 0.233 0.661 0.819 2.319 Silt 0% 26 0.95 0.369 0.122 0.346 0.388 1.098 Silt 2% 25 6.06 0.378 0.129 0.366 0.544 1.540 Silt 5% 25 2.92 0.529 0.155 0.437 0.408 1.154 Silt 8% 25 8.93 0.379 0.195 0.551 0.424 1.199 Sand 0% 25 0.32 0.140 0.052 0.147 0.149 0.422 Sand 2% 26 5.55 0.362 0.219 0.621 0.363 1.027 Sand 5% 25 2.43 0.292 0.430 1.216 0.555 1.570 Sand 8% 26 8.59 0.631 0.396 1.120 0.741 2.097 Table 3-2: Summary of Statistics of organic content measurements after subtracting the inherent organic content Soil Type Source-Design No. of Labs Average Sx Repeatability Reproducibility 1s (Sr) D2s 1s (SR) D2s Adj. Clay 2% 26 2.25 0.505 0.282 0.798 0.576 1.630 Adj. Clay 5% 25 5.32 0.498 0.246 0.697 0.554 1.567 Adj. Clay 8% 24 8.28 0.519 0.232 0.655 0.566 1.602 Adj. Silt 2% 25 1.97 0.313 0.129 0.366 0.338 0.956 Adj. Silt 5% 25 5.05 0.262 0.155 0.437 0.302 0.856 Adj. Silt 8% 26 7.93 0.368 0.196 0.556 0.415 1.176 Adj. Sand 2% 25 2.07 0.262 0.216 0.610 0.337 0.953 Adj. Sand 5% 25 5.14 0.534 0.397 1.124 0.660 1.869 Adj. Sand 8% 25 8.24 0.683 0.372 1.054 0.774 2.192
12 3.4 Statistical Tests for Significance Tests of statistical significance on the ILS data were performed using t-test and F-test. All t-tests assumed an independent one-sample t distribution for 1% level of significance. The t- test was to determine if the difference in the average adjusted mass percentage of burned organic material were statistically significant from the added percentage of organic material. The F-test was to determine if Sr and SR precision estimates of the properties for different soil types with various organic material percentages were significantly different. The results of the tests for statistical significance on the averages and standard deviations of the measured organic content are discussed in the following sections. 3.4.1 Comparison of the Measured and Target Averages A t-test was performed for comparison of the average adjusted burned organic mass percentage and the added mass percentage of organic material. The results of the t-test for 1% level of significance are provided in Table 3-3. As observed from the table, for all soil types, the adjusted mass percentages of organic material are statistically the same as the amount of the organic material added to the soils. It is indicated from the table that the smallest rejection probability values correspond to the clay blends specifying that clay has the highest inherent organic content compared to the other two soil types. Table 3-3: Results of t-test for comparison of the average measured mass percentages with the added mass percentages of the organic material Compare Degrees of Freedom Critical t Computed t Rejection Probabilities Decision 2% Clay 26 2.479 1.974 0.0591 Accept 5% Clay 26 2.479 2.376 0.0252 Accept 8% Clay 26 2.479 2.329 0.0279 Accept 2% Silt 26 2.479 -0.264 0.7939 Accept 5% Silt 26 2.479 1.118 0.2738 Accept 8% Silt 26 2.479 -0.933 0.3594 Accept 2% Sand 25 2.485 0.715 0.4812 Accept 5% Sand 26 2.479 1.380 0.1793 Accept 8% Sand 26 2.479 1.632 0.1147 Accept 3.4.2 Comparison of the Standard Deviations of Measurements The F-test was performed to determine if Sr and SR precision estimates of the properties for different organic contents were significantly different. The standard deviations that are not significantly different would be pooled in development of the precision estimates. The F-test was conducted on standard deviations of different organic contents of the three soil types at 1% level of significance. A rejection probably value smaller than 0.01 would indicate that the differences between standard deviations are significant. The results of the tests for statistical
13 significance are shown in Table 3-4 through Table 3-10 and are discussed in the following sections. 3.4.2.1 Comparison of the Standard Deviations of Clay Blends The results of F-test on within and between-laboratory standard deviations of organic content measurements of clay blends are shown in Table 3-4 and Table 3-5. As shown in the tables, the rejection probabilities for comparison of both within and between-laboratory variability of the clay samples are all greater than 0.01 indicating that their standard deviations are the same and therefore, they can be combined. Table 3-4: Results of F-test on comparison of within laboratory variability of measurements of clay for 1% level of significance Compare Within Standard Deviation Degrees of Freedom Critical F Computed F (Sr) Rejection Probability Decision 2% vs. 5% 0.282 vs. 0.246 25 & 24 2.64 1.311 0.2550 Accept 2% vs. 8% 0.282 vs. 0.232 25 & 23 2.69 1.482 0.1732 Accept 5% vs. 8% 0.246 vs. 0.232 24 & 23 2.70 1.131 0.3851 Accept Table 3-5: Results of F-test on comparison of between-laboratory variability of measurements on clay for 1% level of significance Compare Between Standard Deviation Degrees of Freedom Critical F Computed F (SR) Rejection Probability Decision 2% vs. 5% 0.576 vs. 0.554 25 & 24 2.64 1.083 0.4237 Accept 2% vs. 8% 0.576 vs. 0.566 25 & 23 2.69 1.036 0.4681 Accept 8% vs. 5% 0.566 vs. 0.554 23 & 24 2.676 1.045 0.4568 Accept 3.4.2.2 Comparison of the Standard Deviations of Silt Blends The results of F-test on within and between-laboratory standard deviations of organic content measurements of silt blends are shown in Table 3-6 and Table 3-7. As shown in the tables, the rejection probabilities for comparison of both within and between-laboratory variability of the silt samples are all greater than 0.01 indicating that their standard deviations are the same and they can be combined. Table 3-6- Results of F-test on comparison of within laboratory variability of measurements on silt for 1% level of significance Compare Within Standard Deviation Degrees of Freedom Critical F Computed F (Sr) Rejection Probability Decision 5% vs. 2% 0.155 vs. 0.129 24 & 24 2.66 1.430 0.1936 Accept 8% vs. 2% 0.196 vs. 0.129 25 & 24 2.64 2.310 0.022 Accept 8% vs. 5% 0.196 vs. 0.155 25 & 24 2.64 1.615 0.1222 Accept
14 Table 3-7: Results of F-test on comparison of between-laboratory variability of measurements on silt for 1% level of significance Compare Between Standard Deviation Degrees of Freedom Critical F Computed F (SR) Rejection Probability Decision 2% vs. 5% 0.338 vs. 0.302 24 & 24 2.66 1.247 0.2965 Accept 8% vs. 2% 0.415 vs. 0.338 25 & 24 2.64 1.514 0.1567 Accept 8% vs. 5% 0.415 vs. 0.302 25 & 24 2.64 1.888 0.0620 Accept 3.4.2.1 Comparison of the Standard Deviations of Sand Blends The results of F-test on within and between-laboratory standard deviations of organic content measurements of sand blends are shown in Table 3-8 and Table 3-9. As shown in the tables, the rejection probabilities of two out of three comparisons for each within and between- laboratory variability are smaller than 0.01 indicating that the variability of measurements on sand blends with 5% and 8% organic material are significantly larger than those of the blend with 2% organic material. It is speculated that the non-adhesive nature of sand has contributed to the increase in measurement variability. The more walnut shell grits were added to the sand particles, the more non-adhered organic material were available to segregate. Due to the significant difference in variability values, the standard deviations of the sand blend with 2% organic material would not be combined with those of the sand blend with 5% and 8% organic material. Table 3-8: Results of F-test on comparison of within-laboratory variability of measurements on sand for 1% level of significance Compare Within Standard Deviation Degrees of Freedom Critical F Computed F (Sr) Rejection Probability Decision 5% vs. 2% 0.397 vs. 0.216 24 & 24 2.66 4.006 0.0020 Reject 8% vs. 2% 0.372 vs. 0.216 24 & 24 2.66 3.399 0.0048 Reject 5% vs. 8% 0.397 vs. 0.372 24 & 24 2.66 1.179 0.3770 Accept Table 3-9: Results of F-test on comparison of between-laboratory variability of measurements on sand for 1% level of significance Compare Between Standard Deviation Degrees of Freedom Critical F Computed F (SR) Rejection Probability Decision 5% vs. 2% 0.660 vs. 0.337 24 & 24 2.66 2.474 0.0008 Reject 8% vs. 2% 0.774 vs. 0.337 24 & 24 2.66 4.911 0.0001 Reject 8% vs. 5% 0.774 vs. 0.660 24 & 24 2.66 1.985 0.2205 Accept 3.5 Combining of Similar Standard Deviations The precision estimates for the organic content measurements are computed after combining the standard deviations that were not significantly different. As mentioned previously, repeatability and reproducibility standard deviations corresponding to clay or silt blends with
15 different percentages of organic material were not significantly different. In this respect, the variability values were combined as presented in Table 3-10. For the sand blend, because of the significant difference between variability of measurements, the statistics of the blend with 2% organic material could not be combined with those of the blends with 5% and 8% organic material. Since in the field the inherent amount of organic material in sand is typically low (<2%), the standard deviations of the sand blends with 2 % organic material was used as the repeatability and reproducibility standard deviations for the sand blend materials as presented in Table 3-10. Table 3-10: Combined standard deviations of the soil blends with various organic contents Blend Type Repeatability std (%) Reproducibility std (%) Clay 0.25 0.57 Silt 0.16 0.35 Sand 0.21 0.35 To examine if the standard deviations can be further combined, an F-test was conducted to examine the significance of the difference between variability of various soil types. The results of statistical F test for 1% level of significance are provided in Table 3-11. As shown in the table, for the clay, the repeatability of measurements is significantly different from that of the silt blend and its reproducibility is significantly different from that of both silt and clay blends. Nevertheless, the repeatability and reproducibility of the silt and sand blends are the same. Therefore, the variability of clay should be presented separately from that of silt and clay, while the standard deviations of silt and sand could be combined. Table 3-11: Results of F test for comparison of standard deviations of organic content measurements of various soil blends Compare Degree of Freedom Critical F Repeatability Repro ducibility Computed F(Sr) Rejection Probability (Sr) Decision computed F(SR) Rejection Probability (SR) Decision Clay vs. silt 74 & 74 1.72 2.45 <0.0001 Reject 2.54 <0.0001 Reject Clay vs. sand 74 & 24 2.37 1.40 0.1786 Accept 2.62 0.0050 Reject Silt vs. Sand 24 & 74 2.05 1.75 0.0355 Accept 1.03 0.4422 Accept 3.6 Precision Estimates of AASHTO T267 In developing the precision estimates for AASHTO T267, the standard deviations corresponding to the soil blends with similar variability would be combined and for those with different variability would be used separately. In this respect, the repeatability and reproducibility standard deviations of silt and sand, which were not significantly different, were combined. However, the standard deviations of clay, which were significantly different from those of silt and clay were not combined. Table 3-12 presents single operator and multi-
16 laboratory estimate of variability (1s) and the allowable difference between two results (d2s) for organic content measurements of the soil blends. A proposed precision statement for NCHRP T267 based on the precision estimates in Table 3-12 is provided in Appendix E. Table 3-12: Precision estimates for measurement of organic content of soil Condition of Test and Test Property Standard deviation, % (1s) Acceptable Range of Two Results, % (d2s) Single-Operator Precision: Clay 0.25 0.72 Silt and Sand 0.19 0.54 Multilaboratory Precision: Clay 0.57 1.60 Silt and Sand 0.35 1.00
