03.07.2013 Views

Nested Designs - Scholar

Nested Designs - Scholar

Nested Designs - Scholar

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

3. F-Tests<br />

F = MSA/MSB tests the null hypothesis that the mean transpiration rate is the same for all five<br />

hybrids versus the alternative that the mean transpiration rate is different for some hybrids.<br />

Alternatively, this could be stated as: F = MSA/MSB tests the null hypothesis (7) that the effects<br />

of all of the hybrids on mean transpiration rate are zero, versus the alternative that the effects of<br />

some of the hybrids are non-zero.<br />

F = MSB/MSC test the null hypothesis that there is no variation in mean transpiration rate<br />

among pots within each hybrid, versus the alternative that there is some variation among pots<br />

within some hybrids:<br />

2<br />

2<br />

H0: 0 , versus HA: 0<br />

(9)<br />

B<br />

B<br />

Finally, F = MSC/MSE tests the null hypothesis that there is no variation in mean transpiration<br />

rate among plants within each pot, versus the alternative that there is some variation among<br />

plants within some pots:<br />

2<br />

2<br />

H0: 0,<br />

versus HA: 0<br />

(8)<br />

C<br />

C<br />

As explained above, the latter two sets of hypotheses are stated in terms of "variation" rather than<br />

in terms of "means" or of "effects" because the effects of pots and plants are assumed to be<br />

random variables rather than fixed constants. Since the effects are random, we cannot assume<br />

they sum to zero as we would if they were fixed. Therefore it would be meaningless to test the<br />

hypothesis that all the random effects are zero. On the other hand, the assumption that the effects<br />

are random variables does make it reasonable to test the hypothesis that the variance of those<br />

random variables is zero, and that is what we are doing. An effect is random rather than fixed<br />

when the levels tested are a random sample from the set of all levels of interest.<br />

nested01.docx 10 4/5/2012

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!