Is it true that ...

Can’t I just choose where to put the plot?
Really – my judgment is very good!

This is good question. The simple answer is "No" – or more likely "good grief NO!!!," the actual answer is "Yes," if you promise not to use the data in certain ways (and keep your promise).

The principle to bear in mind here is that the final answer must come from a real sample, and that means that you do not adjust those sample locations. You might put more plots into some strata than others, but the placement inside that strata must be appropriate (systematic or random).

On the other hand, if the plots are an intermediate item you can do anything you want. (Actually, if you are risking your own money you can always do anything you want to. Statistics and sampling are a safeguard, and you can always discard the safety net if you are the one that takes the fall. (The spectators, of course, are always willing to have the net removed).

Consider an example:

A photointerpreter is going to put in a set of photo plots to give a relative volume to each stand. A computer may measure the trees or do some other magic, but the person gets to choose the place in the stand to put the plot.

These estimates are then going to be corrected by a set of plots put into the stand in a random or systematic way in order to remove any bias in the original photo estimates.

In this case, the photo person can choose "typical areas" in which to put the photo plots. The distinction here is that these plots will not be directly used to establish the final volume. Think about the units of the final answer. Where do they come from? They come from the ground plots. The photo plots can be in some screwball unit like "chunks/plot." The ground plots will be correcting these units by establishing the relationship of "BF/acre for every chunk/plot."

Now the photo person is never going to choose a road or small clearing to put a photo plot into. That is perfectly OK, because the ground plots will fall into these kinds of places with approximately the right frequency.

Consider a second example:

An outfit is trying to balance their machine use. They have a few dozen stands, and want to get the best answer in each of them. They do not mind if the overall answer is biased, since they are just interested in the relative volumes in each stand in order to allocate machines and staff.

In such a case, there is nothing wrong with choosing where to put the plots, based on the cruiser’s judgment. When time is short and samples are small, such a technique is often superior to valid samples. Give the measured volume in a silly unit like "yams/acre," if possible, to remind everyone that the actual number is not to be used. As a relative measure, it may be quite adequate.

(I know that using a silly unit sounds awkward, and not "serious" enough but if you put this bogus data in "BF/acre" you can be sure some well intentioned ninny will misuse it. Do it only under protest – but even an attempt will make the point. If they want real units, they should have to put in real samples.)

Consider a third example:

A photointerpreter is trying to "calibrate" themselves by comparing the ground situation to the photos they are using. They would like to look at a part of the stand that is "typical" of the large stand, but in some small area they can reach easily.

There is nothing wrong with visiting areas chosen for convenience and typical form. If they insist on gathering data, then they should estimate everything in order to emphasize that it is not a "serious measurement" that could be used to get total volumes.

All of these situations, and many others that we might encounter, are examples where the totals generated by the plots are not used to determine the total volume (at least not directly). They are used as intermediate values, or to establish relative amounts with small sample sizes.

To keep it simple, we might say "YES, you can choose the sample location, but NO, you cannot use the results." Put those results into a silly unit to keep yourself from thinking that it is "real sample data," but if it suits your needs then choosing the sample location can be an appropriate method in some situations.