Hi Gurmeet,
I'm going to reply to the list so others can add their opinions. Since you have pre-flood data too, I think that what you want to do is actually much simpler than we talked about first. It sounds like you just want to compare the top layer before the flood with the top layer after the flood. Is that correct? If it is, then you basically have a simple one-way ANOVA or a t-test. Your response is the measured soil variable in the top layer, and your treatment is pre or post flood. Since you're using different soil cores pre- and post-flood, your samples are independent, so you don't have to get involved in mixed models. If instead, you want to make comparisons within cores, then you need a mixed model. Jeff From: Gurmeet Singh [mailto:[email protected]] Sent: Thursday, November 15, 2012 6:43 AM To: [email protected] Subject: Re: [ECOLOG-L] Suggestions needed -ANOVA Dear Jeff, Thanks a lot for the reference , The paper seems useful. Yes , I am planning to test the hypothesis that the top layer of sediment is compositionally different from lower layers after the flood? Geochemical Analysis has shown beautiful results and I wish to prove them statistically . I do have the background values for the area as a long term study has been going on. Shall I go for linear mixing models ? Regards Gurmeet ====================== On Wed, Nov 14, 2012 at 9:07 PM, Jeffrey Evans <[email protected]> wrote: Hi Jawaharlal, It's not totally clear to me from your description what you want to compare. * Are you proposing to test the hypothesis that the top layer of sediment is compositionally different from lower layers after the flood? * Do you have any control samples from areas that weren't flooded? * What is the measured response variable? If you want to make comparisons within individual cores, you will need to control statistically for core and should use a mixed model with a random effect for core. Read Ben Bolker et al.'s 2009 paper (below). However, if you use sample depth as a categorical predictor and you have 5 depths this will just be a test of whether there are differences in the mean response variable across depths. Is that the answer you want? That's vague advice, but I'm not sure I have enough information to say anything more specific. Best, Jeff Evans Bolker's paper: Bolker, B. M., M. E. Brooks, C. J. Clark, S. W. Geange, J. R. Poulsen, M. H. H. Stevens, and J.-S. S. White. 2009. Generalized linear mixed models: a practical guide for ecology and evolution. Trends in Ecology & Evolution 24:127-135. -----Original Message----- From: Ecological Society of America: grants, jobs, news [mailto:[email protected]] On Behalf Of Gurmeet Singh Sent: Wednesday, November 14, 2012 4:03 AM To: [email protected] Subject: [ECOLOG-L] Suggestions needed -ANOVA Dear ECOLOG member *I urgently require your suggestions related to statistical approaches to the following problem* *The Problem* I have been working on a coastal forest ecosystem and its sedimentary biogeochemical processes. Recently due to cyclonic event, the area was suddenly flash-flooded. The flood supposedly disturbed the sediment column as well as retreating water deposited waste load . *Approach* Sediment cores (n=10) , collected from the area immediately after the flood and sectioned at 5 cm interval for the geochemical analysis. *Hypothesis* *The flood has resulted in significant changes in the sediment column biogeochemistry .* The hypothesis is well supported by the experimental results . Geochemical analysis of the cores at 5 cm interval indicates trends of deposition of fresh sediments/disturbance/waste deposition *Statistical approaches* 1. .Entire forest area cover is disturbed due to cyclonic flood i.e. the spatial variation in the geochemical characteristics are not significant 2. The sediment columns has been disturb due to deposition I.e variation in the geochemical characteristics along with depth are significant . *Help Needed* * * *Which test of significance will be more suitable. * *I have applied ANOVA but have been getting conflicting response that * * * *" This is not totally appropriate for the samples collected especially in sediment cores. Certain geochemical processes occur **down a sediment column and there could be relations between samples. Therefore, the samples (from different depth of a core) should not be treated as independent samples and should not analyzed by using ANOVA ."* Indeed, I have treated the different sections of a core along with depth as individual independent samples and have compared with the correspond values of other cores at any given depth. *Now I am confused whether my approach is wrong or ANOVA can't be applied to the system ? What other statistical approached (apart from PCA) I can do to check the significance of variation across depth* * * Any response in this regards will be highly appreciated . If you can forward me the some of related research papers, I will be obliged . Received response will be duely credited . Regards Dr. Gurmeet Singh Jawaharlal Nehru University New Delhi ======================
