INTRODUCTION
On Friday, February 9th, an area was sampled randomly and judgmentally in order to find the twenty sites in the sample universe. The North and South had mountain topography while the middle of the area had a river in the shape of a Y stretching from the West to the East. The river was surrounded by a floodplain and the base of the mountain regions had foothills as well as a section to the West between the two branches of the river once outside the floodplains.
PROCEDURE
Using a number generator, 20 squares from the site universe were chosen and marked. Kacie wrote down the list of squares selected and inquired about sites within while I marked the squares chosen on the map and the sites found. Our group noted that the random sample found four small sites and one medium site, mainly in the area surrounding the river. Since the sites were not very varied in size and were spread across the sample universe in location, Camron suggested selecting one in every five squares, not repeating random sample squares in order to give each topographic type a chance at selection. Erin selected 81 as the starting point because our group hypothesized that the squares near the river bank would more likely contain a site. Kacie wrote down the numbers again and inquired about sites while I marked the squares chosen on the map and the sites found. Our group noted that the judgmental sample found three medium sites and one small site. This result added more depth to the previous understanding of the site universe but no large sites were located which limits the usefulness of the samples. After these two samples were charted, our group discussed the results.
DATA AND RESULTS
At the end of both sample selections, 40 out of 100 squares had been tested and 45% of sites located. While none of the sites found were large, having near half of the sites recorded is enough to start testing hypotheses and drawing conclusions about the sample universe. With five small sites and four medium sites found out of a total of twenty sites of which nine are small, six medium, and five large, developed the sampling understanding of the sample universe. Our sampling efforts lead to a positive understanding of medium sites and their distribution and a fair understanding of the small sites given the amount of each found. However, since no large sites were discovered, how the large sites relate to the other sites as well as their locations and distributions remained unknown information.
The random sample found five of the sites; four small and one medium in size. Since the random sample used a number generator to get rid of bias, variability and pattern was left to the unknown as well. While the random sampling found five of the twenty sites, not many of the sample locations ended up being close to actual site locations and because each location has a 1% probability of being selected, there is little to guarantee finding sites.
The sample universe has a varied terrain and the sites found reflect that. While many sites from the random sample were found by the water source, these sites remained on the smaller size and could not properly reflect a dispersion of the sites, especially those of larger sizes. Based on what was gathered from the random sample, it would appear that the area is compromised of small sites near water and at the mountain base. The sample universe contains sites at varied locations and using a strongly methodical process without factoring in other elements of the sample universe such as topography can lead to a poor sampling of the region. The random sample views each of the squares equally regardless of location or probability within each topographic zone: the fact that the floodplains would have a higher chance of containing a site compared with the mountain region is simply not considered. That being said, the sites were varied across the sample universe and giving each square an equal chance at selection allowed for some of this variation to be expressed. Our random sample located sites in the floodplains and foothills in different regions of the map which was an indicator for location variation.
The judgmental sample found one fewer site compared to the random sample. With the random sample, out of 100 squares, each square had a 1% chance of being selected and a 20% chance of containing a site. There was a 5% chance that each selected square would contain a site. With the judgmental sample, the probability of a square becoming selected relied upon the other squares and their relation. Our group used a one in every five square method, modified slightly in order to not repeat squares from the random sample. In this way, the probability of a square containing a site was dependent upon the other squares selected. Using this method, the first selected square shifted the probabilities of the squares that were selected and the ones that were not. This method also made an assumption about the spacing and diversity of site settings which impacted the ability to locate squares with sites. Our judgmental sample was not more effective than the random sample but a different method could have been. A lot of the sites seem to be along borders of the different topographic zones: flood plains to foothills, foothills to mountains, river banks, and so forth. Concentrating a sample on squares that these border take place at would definitely have been more effective in locating sites. With twenty squares containing a site in the sample universe, variability is almost guaranteed and implementing a method that accounts for such is important. That being said, the sample universe shows certain trends with its samples, making it important to hypothesize potential patterns.
Selecting 20 out of 100 squares does not cover enough ground to give an accurate representation of the site selection. I would say that 33% to 50% would give a better understanding of the sample universe without being inefficient or sampling too many squares. Sampling all of the squares would be a huge waste of time and resources especially given that only 20% of the sample universe contains a site.
CONCLUSION
CONCLUSION
There are a multitude of methods for sampling a site with their own risks and benefits. Random sampling is a great starting sampling method since it does not make any assumptions on the sample universe before data is collected. However, random sampling treats every square with equal risk which can impact finding sites while sampling. Using judgmental sampling requires tact for the reasoning for selecting squares is much more individualized, creating higher risk. When it comes to archaeology, it appears the best method is oftentimes trial and error.
No comments:
Post a Comment