# Questions from the Field ...

### "How many random numbers do I need to pick a random tree on my plot?"

This is a really good question. It’s from an organization that goes to the trouble to choose random trees on some of their sample plots in order to check bark thickness, radial growth, etc.

Now these data are often available from trees you bore for age or for measuring site index, but those trees are not a valid sample of all the trees in the inventory and they do not know if it is close enough just to use the borings from this selected group (note – deep belief that it is OK is not a substitute for knowing).

These folks are making the effort to check. Smart move. If the results come out the same they can use the increment bored trees (past and present) without worrying, and if there is a difference they know how much to adjust the bored information.

Rather than using the "first tree from North" or some other biased scheme they did it right. First they painted numbers on the trees, starting from North in a clockwise direction. That way the check cruiser could see that they had no option to control the choice.

The crew then chose the tree with a random number list that was sent out with the crew for each plot, so the crew could never be accused of cheating. Note: this is for the protection of the crew, the people who do the work – nice touch, don’t you think?

They only wanted to randomly select from live trees, so they could not send the crews out on the plot with only one random number. Perhaps only trees 3 and 5 were alive, so unless a 3 or 5 existed in their short list they were stuck. Even with a longer list of numbers, it is possible that neither number would appear.

The Solution:

Send along a list as long as the number of trees might be (perhaps 16 might be the maximum). Rather than send out just random numbers, take the numbers 1 – 16 and sort them into random order. In the field, go down the list until one of the numbered trees qualifies for selection and that is a random choice from the available trees.

Sorting a list into random order is easy with most spreadsheets. List the tree numbers 1-16 in one column, generate random numbers into the next column (and "fix" them so they do not change). Now sort both columns based on the random number column. The tree numbers are now in a random order.

Print them out, with the plot number on the top of the list, and the crew will always be assured of getting a random selection with that short list. In addition, anyone can trace the process and convince themselves that it is carried out correctly in the field.

This outfit has another good idea. If the processes designed in the office do not make sense in the woods the cruiser has the option of declaring a "cruiser emergency" and including the added data they think is appropriate – for instance a different tree for site index. Over time they find out how often the cruisers do not think the processes worked correctly, and in the end they can see how much difference it would have made to do it the way the cruisers would have suggested. A good system for all parties.

Originally published July 1999

Back to Contents