APPENDIX B:Sampling Guidelines
The sampling strategy outlined below has two separate, but complementary goals: (1) estimating the artifact density of a scatter, and (2) estimating the artifact attribute population parameters. Refuse scatters will be sampled only when it is not feasible to analyze all artifacts; for lithics this is more than 100 items, for ceramics, 200. The goal of sampling is to obtain a reasonable estimate of the scatter population parameters, which requires a representative sample. More often than not, simple random sampling does not result in a representative sample. A more representative sample may be obtained by stratifying the scatter into quadrants, gridding the scatter within each quad, then randomly selecting a given number of grid units within each quad. A third concern in the sampling and analysis process is to maintain the original character of the site by returning artifacts as close as possible to the locations they came from.
1. Define Scatter, Determine Scatter Area, Identify Quadrants. If a scatter is to be sampled, define the scatter boundaries, measure the scatter (length, width), determine scatter area, and divide the scatter into four roughly equal quarters. 2. Determine Size and Number of Sample Units. The size and number of sample units will be determined by the total
scatter area. The total area sampled should equal 10 to 15% of the total
area of the scatter. Size of units should range from 2x2m to 10x10m. All
sample units for a single scatter must be the same size. Keep in mind
that a large number of sample units is analytically preferable, but may
be impractical. Select the size and number of these units weighing the
costs and benefits of both these concerns. For example if you have a
scatter that is 4500 m 3. Select Sample Units. Use the bottom of the density grid sheet or a separate sheet of graph
paper to outline the scatter boundaries and superimpose the grid and
quads. Number all grid units that appear to have at least half their
area within the scatter. Select sample units from each quad using a
random numbers table until you have chosen units equalling 10-15% of the
surface area of the entire scatter. The number of units selected should
at a minimum include 4 units (1 per quad); at a maximum, 20 units (5 per
quad) for the entire scatter. As noted, the number of sample units
should be 4. Locate Sample Units and Record Artifact Counts. Sample units can be located in two ways: 1) establish a density datum (on the ground and on your map); using your map, establish the distance and azimuth to your grid unit (pick a corner) from the datum; or if using dogleash units, establish the distance and azimuth to the center; then using those coordinates, locate that point on the ground by shooting the bearing and measuring the distance from the datum. OR 2) using the grid system you have created on your map, figure the X, Y coordinates of your selected sample units; then using the metric tapes and pinflags, project the Y axis on the ground and at the appropriate intervals shoot a 90 bearing and measure out to each calculated X coordinate. Once you have located the corner or center point, lay out and label your unit, and pick up and count all artifacts in it. Label the bag with the sample unit number. Record counts on the density grid page. (If the number of items in your sample units is highly variable, and you have a relatively small number of units, you will probably want to select additional units. Additional sample units will help to decrease the population variance, and thereby allow a more precise estimate). 5. Return Artifacts to Sample Units. After all analyses are completed, return all collected artifacts to the sample unit from which they came. This is essential to preserve site integrity.
Sampling for artifact analysis will be carried out within the framework created by the artifact density sampling. Our experience last year showed that ceramics rarely required sampling, but lithics were frequently sampled. The following discussion applies primarily to lithics, but the same principles can be used for ceramics when necessary. The goal in lithic artifact analysis sampling is to analyze 30-40 out of 100 artifacts picked up from the four quads of the scatter. The procedure by which the 100 artifacts are selected is outlined below. 1. The sample units to be included in artifact analysis are selected randomly, in a "round-robin" manner from all four quads (these units will in most cases be the same as those established for the density grids). One sample unit is selected (using random numbers) from each quad until approximately 100 items have been picked up. After the first round (one sample unit from each quad), the number of items included is counted. If more than 75 items are selected, then no additional samples need be selected. If fewer than 75 are picked up in the first round, then you will need to select 4 more sample units - one from each quad - until at least 75 artifacts have been picked up (although we are aiming for 100 lithics, having to pick up 4 units at a time may result in picking up more than a 100 lithics - this is okay, but the sample should be designed to stay below 200 lithics, see "sampling strategy for very dense scatters" below). No more than 300 items (total of 3 samples) are to be included regardless of scatter artifact count. 2. For scatters where the number of items picked up is less than 200, select one simple random sample of 30 items for analysis using the random numbers board and table, doubling up artifacts on the board if necessary. Analyze artifacts as outlined in lithic and ceramic manuals. 3. If the estimated number of items exceeds 20,000, then 15 items from each quad must be analyzed (total of 60 items analyzed). The maximum number of items to be analyzed is not to exceed 90. This would only occur if the total estimated was 30,000 or greater. On scatters of this artifact density, an equal number of items from each quad should be analyzed in order to ensure that the analysis sample includes items from all quads of the scatter. This will also divide the sample into units that are of more manageable size to be laid out on the random numbers board.
If after laying out the density grid as described in
the sampling directions it appears that you may pick up more than 200
items in each density grid unit, then the density grid unit should be
subdivided into a smaller grid unit for artifact analysis sampling. The
goal is to keep the size of the sample unit consistent among quads and
avoid picking up more than 600 items. It does not matter if the grid
size for artifact analysis is not the same as the grid size for artifact
density estimation. For example, if you have a very dense scatter
(i.e., 2 items/m
As with the sampling strategy outlined above you are aiming for approximately 100 flaked stone items, or 1% if estimated artifact count is greater than 10,000 items with no more than 300 items (3 samples) included regardless of scatter artifact count. Unlike the refuse scatters associated with structures, we will not be establishing a grid for the nonstructural scatters since we are not doing density estimates at these scatters. Because of this, the artifact analysis units can be placed judgmentally. We still want multiple samples to insure a representative sample of all parts of the scatter (i.e., four 5x5), however these sample units can be placed judgmentally - in areas that would best and most efficiently provide a representative sample of the scatter.
When doing the artifact analysis at scatters that do not require sampling (fewer than 100 flaked lithics and 200 ceramics) a major concern is to maintain the integrity of the artifact scatter. Artifacts must be put back in their original position. This can be easily accomplished by marking artifacts with numbered flags (one set of numbers for lithics, one set for ceramics). For lithics, use the random numbers board to select the 30 items to be analyzed, find the corresponding numbered flags then record those lithics right at their location. All groundstone in a scatter needs to be recorded and this is most easily accomplished by moving from item to item. For ceramics, we found it was easiest to lightly write the number of the flag on each artifact (use pencil), bring all the sherds together for analysis, then return each to the appropriate flag.
peco/cris/appb.htm Last Updated: 13-Feb-2006 |