I have an interesting experimental design. There are 2 treatments (fertilize and non-fertilize). There are 3 plant species. For each species, they are planted on a strip of land which half is fertilized and the other is not. The experiment was conducted
with 3 replications. We would like to test if there is a statistically significant difference between treatments and if there is a statistically significant difference between species. But, after the fact, we found out that for one rep the strips were
fertilized before and the others were not. This has impacted the results. I am looking for a second opinion and if a block effect will be helpful in removing the effect of fertilizer application before the experiment. I hope this makes sense.

Thanks!

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On Wed, 10 Feb 2016 05:59:53 -0800 (PST), "Ilovestats!!" <lucia.costanzo47@gmail.com> wrote:

I have an interesting experimental design. There are 2 treatments (fertilize and non-fertilize). There are 3 plant species. For each species, they are planted on a strip of land which half is fertilized and the other is not. The experiment was conducted

with 3 replications. We would like to test if there is a statistically significant difference between treatments and if there is a statistically significant difference between species. But, after the fact, we found out that for one rep the strips were
fertilized before and the others were not. This has impacted the results. I am looking for a second opinion and if a block effect will be helpful in removing the effect of fertilizer application before the experiment. I hope this makes sense.


Keep in mind that I have no experience with agricultural data,
and that I do not know what /you/ refer to when you ask
about "a block effect" -- The best that I guess about that is,
"No, a single d.f. contrast would not account for interaction"
that might be present.

It looks to me like your robust analysis must be satisfied with
analyzing only the two replications that met the design.
With an N this small, you must be looking for large effects that
you want to estimate the size of; an interaction would make
the estimates and tests inherently challangeable. Now, if
you get the same effect by putting in the 1 d.f. contrast, you
can report that "more complete" analysis, too, noting that the
failure of the design did not turn out to be important, after all.


--
Rich Ulrich

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

I am a little puzzled.

Are you saying that there was no fertilizer at all in two of the reps? Or was fertilizer applied after planting in two reps and before planting in one rep?

Or was one rep (only) fertilized before planting and all three reps were fertilized after planting?

Or, is the real problem that both treatments in one rep were fertilized but only one treatment (fertilized) in the other reps?

Are these species quite similar, eg, cabbages and brussels sprouts, or are they quite different, eg, soybeans and winter wheat? Or, more cogently, why would a conclusion about fertilization of different species apply equally to all of them?

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Hye, Im Nadia..new to this group.. I would like to discuss about experimental design with focus on factorial design. Is anyone can discuss with me about the most recently topic for experimental design especially for the test for normality, variance
constant and independent data. Thank you.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)