Laserfiche WebLink
Corey Heaps <br />CAM Colorado LLC <br />June 21, 2011 P age 1 9 <br />area and the undisturbed area. Both random and systematic sampling designs are <br />acceptable." The vegetation Guideline on page 5 states "it is important that unbiased <br />samples be obtained. In general, the Division recommends that transect samples be <br />randomly located." By definition, randomization means that every individual in the <br />population has an "equal chance of being selected" with the sample area. The DRMS <br />comments are confusing in that they suggest that unless "sequentially numbered transects" <br />are sampled there will be bias in the sampling effort. <br />It would appear that the comments contained in the contained in the Division's adequacy <br />response letter are inconsistent with the regulations, vegetation Guideline and definition of <br />"randomization" as defined in numerous statistical references examined. In a classic <br />statistical treatise, written by D. R. Cox entitled "Planning of Experiments" published by <br />John Wiley & Sons in 1958, he defines "randomness" as consisting of a situation where <br />"there is no recognizable pattern." He further states on page 27 that "the order in which <br />plots [transects] are cultivated or harvested would ordinarily be assumed negligible ... and <br />that the best procedure ... to test. . . from one block [is] in random order ... " <br />Thus, it is irrelevant, if the transects are sampled in ascending or descending order or any <br />other random order. This text further states [page 71] that one should "choose a starting <br />point in a haphazard way without looking at the [random number] tables" while the DRMS <br />herein appears to be saying that random numbers can only be used if a "systematic <br />sampling" approach is used by starting only at row one, column one and then only if <br />random numbers are collected from "sequentially numbered" rows and columns. This <br />approach is contrary to the definition of "random" and Cox's suggestion that the data can <br />be collected in any order as long as the starting point is chosen in "a haphazard way." This <br />text spends several pages discussing the "drawbacks of systematic" sampling schemes and <br />concludes by saying [page 81 ] that "subjective allocation of treatments to units [sample <br />transect location] should never be used, because the method has serious disadvantages and <br />no compensating advantages when compared to objective randomization." Cox states <br />[page 85] that "it is not that the systematic arrangement is necessarily less precise than the <br />randomized one, but that the assessment of the results is on a less objective basis." <br />Numerous other texts contain similar cautions. <br />There is an abundance of documentation clearly showing that a "systematic sampling" <br />approach suggested by the DRMS yields data that is more, not less biased, as are <br />suggested by the Division and the most unbiased data is obtained from a completely <br />randomized sampling approach. By implementing a two -stage "random sampling <br />approach," such as was used herein, by firstly selecting purely random coordinates and then <br />secondly, by sampling those coordinates or transects in a random manner, yields <br />significantly less unbiased data that the "systematic sampling" approach being suggested <br />by the DRMS. <br />The DRMS is herein apparently suggesting that the methodology used to collect the <br />field data is somehow flawed and that documentation must be submitted "demonstrating <br />