spot the flaws in following statistical experimental designs

Question

(i) Researchers were interested in whether soil type affected growth rate of a native Acacia species commonly found growing on sandy soils. They established loam soil plots by digging large holes and filling them with loamy soil. Acacia seedlings were planted into these plots, and into neighbouring undisturbed plots with sandy soil. Growth rates of the seedlings were compared.

What error have the researchers made? How could the researchers improve this experiment?

(ii) Researchers were studying the effect of position on a hillslope (lower slope, mid slope, or upper slope) on regeneration of gum trees. They randomly selected six hillslopes for the study. At each position (lower, mid, upper) on each hillslope, three 50 x 50m quadrats were laid out randomly. Thus, there were 3 x 6 = 18 quadrats at each position. On each quadrat, they counted all the gum tree seedlings. The data were analysed using a one-way ANOVA with position as the factor and n = 18.

What error has the researcher made? How could this error be corrected?

(iii) Researchers carried out an experiment to determine whether male mantids use chemical signals to locate females. Forty male mantids were obtained from the wild. The first twenty obtained had their antennae removed while the second twenty obtained just had their antennae snipped. The mantids were then placed, one at a time, in a Y-shaped chamber with a choice of two compartments: one containing a live female, and the other empty.

What error have the researchers made? How should this experiment have been carried out?

Answer 1

(i) .... and into neighboring undistributed plots with sandy soil

So not all of the soil was sampled. the plots with the sandy soil should be

placed randomly so ALL portions of the soil are tested.

(2)

two of the conditions of the ANOVA are not met; 1) The dependent variable (# of seedlings) is technically

a discrete, not a continuous, random variable; There is perhaps a work-around that will solve this...

2) The dependent variable must be normally distributed... so they are operating under the assumption that the # of seedlings is normally distributed. Again, there is perhaps a work-around , or a preliminary test to verify normal distribution. Finally a sample of 6 hillslopes does not give the experiment

much power. Recall for normal distribution, there should be at least 30, but the more the merrier.

(3) Why did they snip the antennae