Replications and Subsamples Needed to Show Treatment Responses on Forest Soils of the Coastal Plain
1983
Lloyd, F Thomas | McKee, William H.
This research illustrates how randomized, complete-block designed field studies can be insensitive to detecting treatment-caused differences in the concentration of soil nutrients. An example shows that variation without compositing can be so large that 727 blocks are needed to be 90% confident of detecting a difference in total nitrogen of 100 µg g⁻¹. Graphs are presented for determining the optimum numbers of blocks and chemical determinations per plot. The results show that compositing does not affect the number of chemical determinations per plot, but does reduce the number of blocks in direct proportion to the reduction in variance resulting from compositing. A hypothetical example shows that the number of blocks can be reduced from 727 to 15 by constructing blocks with less natural variation between treatment plots and by compositing 20 sample cores for each chemical determination.
显示更多 [+] 显示较少 [-]