How would the researcher apply the scientific method to this problem?

Question

Consider the following hypothetical problem: An agricultural researcher is attempting to produce a variety of corn that will produce 3-5 ears of corn per corn stalk. For this to occur, the researcher knows that the corn will only germinate in a soil composition that will absorb and hold a maximum amount of water. Because of other factors, the researcher is limited to using sand and peat moss. The researcher needs to know the ratio (the amount of each soil type) that will produce these conditions.

Answer 1

So the researcher sets up a series of experiments, each with a different ratio of sand and pet moss from 0:100 throough 10:90 .... to 90:10 and 100:0, then sows a numebr of seeds in each plot and waters as normal.  The ratio which holds the water best will produce the highest yield of the required product.  This can be refined into smaller steps in a second series once the broad range is narrowed down.

Answer 2

OK, this one will be a long answer. First, the researcher would need access to a large sample of seeds to sow, or a large number of seedlings already growing under controlled conditions in, say, a greenhouse.

Additionally, the researcher would need access to a large field (a greenhouse is often preferable for such trials, but corn's a bit big to grow like that).

So, now the researcher would set up a number of plots within the large field. Within each, they would demarcate a number of sub-plots and _randomly_ assign each to a particular ratio of sand to peat moss. To properly account for environmental differences across the large field, one would ideally have more one sub-plot representing each ratio in each plot.

Then, seeds or seedlings would be randomly assigned to each sub-plot, with multiple plants potentially growing in each such sub-plot.

At the end of the experiment, the researcher would calculate the mean number of ears of corn per corn stalk for all the plants in a sub-plot (because these values are likely not to be truly independent of one another). Provided there were sufficient numbers of plots, then the last step would be to subject the data to a split-plot analysis of variance.

For more reading, see:

http://www.tfrec.wsu.edu/ANOVA/RCBsp.htm...

http://www.asu.edu/sas/sasdoc/sashtml/st...