Questions from the Field:
Is there any advantage to using lines of plots (rather than a grid)?
A grid is almost always the best way to lay out a given number of plots (you can always create an exception to this rule, but you would be hard pressed to find one in the field). There are very slight advantages to a “triangular” grid vs. a “square grid”, but let’s ignore that for the moment as being too small an issue to bother with. A grid has several advantages of a non-statistical nature to consider.
First, you cover the ground much better, and have more opportunity to notice anything of interest to a client or your logging group. That appeals to the folks that are counting on you to notice these things on their behalf. Second, just about everyone has a good reaction to a grid, and thinks that it “clearly covers the ground well” and provides a good sample. This kind of emotional credibility has real value. Third, a grid actually is “more systematic”, and therefore more efficient at ensuring that the large and small values average out quickly to give a good average result – particularly when any trends in plot data are not obvious. When any geographic trend of what you are sampling is not easily observable on a photo, a grid is sure to catch it, but a line of plots might miss it, which is one reason that lines are not as efficient as a design.
On the other hand, a line of plots is faster to complete. That makes the cost per plot lower, and that kind of trade-off is often not trivial. The shorter length of ground covered may not matter so much in Eastern Oregon, but on the Oregon coast and other really ugly ground it matters a lot.
The key, however, is that with a bit more time along that shorter distance you can put in extra plots along the lines with very little additional cost (I am thinking principally about count plots here). Of course extra plots can also be installed along the approach to each grid point, but the resulting map does not look as comforting, and technically (and it is a small technical point) is not unbiased because the plots do not then cover the entire area equally. These extra plots along a line can overwhelm the inefficiency of the line layout. The same time or money allows more plots, and therefore a better answer.
Can you beat the psychology, however? Can you handle the “what if you missed something important in that big gap between the lines?” question. Can you satisfy yourself in that regard? Perhaps you can display photos of the area to calm any concerns that something very different exists between the lines of plots. If you can handle these concerns, lines of plots are a perfectly legitimate option for a sampling design.
Originally published March, 2006