New to forum

#1
Hi all,

I'm excited to find this discussion forum. I think it offers a great opportunity for those of us without easy access to a community of statisticians. I am a molecular biologist and hope I can offer some help along the way.

In the meantime, I'm hoping for a bit of help myself. I have a some data that I've never considered before. I've collected data on the distribution of bacteria through a plant, including crown and roots. I have presence/absence as well as abundance of the bacteria at numerous sample points throughout the plant. I further have distance and direction from the trunk for each sample point. It strikes me that these data fall within the realm of circular or spherical statistics, yet I have no experience in these areas. Any suggestions you can offer would be helpful.

In addition to the broader analysis question, I further have some specific issues I'm unsure how to deal with. I have multiple sample points at increasing distance along a single branch (or root) from the trunk. In many ways these sample points are not independent, but in some ways could be. Sample points on a branch that's completely filled with bacteria are likely not independent - the bacteria almost certainly spread from a single source. However, some branches may only have the bacteria at the tip...or in two tips. In this case, the remaining sample points may be independent. Should I consider these sample point not as independent samples, but instead as measures of the extent of infection? That is, each branch would be independent but the samples on each branch would be used to calculate a "distance" of infection? Does any of this make sense?

I'd really appreciate any help this community can offer.
 

noetsi

No cake for spunky
#2
I have not seen the terms circular or spherical statistics used in the literature (other than sphericity in factor analysis which I doubt you are referring to). Is this a specific methodology or is it a substantive concept inside biology? One problem with statistical analysis is that different fields use specific terms to refer to a method that other fields call something else.

To answer your question I think you first have to consider carefully what you want to answer, what is your specific research question and hypothesis. Depending on what it is your question might have different answers. It is very difficult to address generic questions, posters are more likely to respond to narrow, concrete questions in the context of a specific method. I think that you would consider sample points that logically are connected to be dependent and ones that logically not connected as indpendent. So the answer to your question of whether sample points are independent or dependent depends on the type of differences you note. Sometimes they will be independent sometimes not depending on the likely reality and what question you are answering.
 
#3
Thanks for the response and the advice, noetsi. Circular or spherical statistics are branches, I suppose, of directional statistics. Fields that deal with directions and angles, time and cyclical events - designed to deal with the fact that the mean direction between 10 degrees and 350 degrees is not 180 degrees. It's a field applicable biology, physics, chemistry, geology, and geography...but one not often taught (at least in the courses I took). I posted here because, although there are books on the field, I was hoping to get a jump start and save myself wasted time down false avenues.

My goal is to assess variation in distribution and abundance of the bacteria within the root system and within the crown, as well as differences between root and crown.

Should I post these questions in a different forum than the welcome forum? I was hoping by posting here more people would see it and perhaps direct me to those with experience in circular (directional) statistics.
 

noetsi

No cake for spunky
#4
We have a biostatistics forum and I would guess you would find any poster familiar with these approaches there. I don't know how commonly posters actually come to this forum to look at statistical questions. It is primarily intended for introductions. Unfortunately I have never encountered the methods you reference so i am no help.

You will find, I am sure you know this already, that people specialize in certaint type of statistics and it is often difficult even here to find people who specialize in what you are interested in. I come here for time series questions most often, and since those are largely used in econometrics the only ones who are knowledable are econometricians who come and go from this forum. It might take a while for someone who does work in your area to respond.