NPMR vs. SEM

[paish...@gmail.com]

There are so very many statistical methods available that it is often confusing which to use.  I have been using SEM for a few years and often get results that support my model hypotheses. I have had HyperNiche on my computer for a few years, tried it once in a while but never delved into it much.  Are NPMR and SEM comparable enough to not make much difference which I use? I like SEM because it is somewhat straight forward in relation to interaction effects.  But I like the idea that some predictors may have multiplicative interactions and could justify using NPMR.  I work mostly with freshwater aquatic macroinvertebrate and zooplankton so there are many inherent problems associated with these taxa that are always of concern.  Thanks! 

[Bruce McCune]

SEM and NPMR are actually quite different. To name a few differences:

- SEM is based on a linear model; NPMR assumes no particular shape of relationships. (SEM works on covariance matrices while NPMR uses multidimensional smoothers.)

- Interactions must be modeled explicity with SEM but are implicit and automatic with NPMR.

- SEM requires a structured hypothesis on how the variables are interrelated while NPMR starts with no assumed structure Instead, NPMR uses the data to expose relationships of the form y = f(x1, x2, ...) where f is an unspecified smooth function. SEM users usually have much more complicated structures to evaluate.

Hope this helps.

-Bruce McCune

On Tue, Jan 8, 2019 at 3:00 PM <paish...@gmail.com> wrote:

There are so very many statistical methods available that it is often confusing which to use.  I have been using SEM for a few years and often get results that support my model hypotheses. I have had HyperNiche on my computer for a few years, tried it once in a while but never delved into it much.  Are NPMR and SEM comparable enough to not make much difference which I use? I like SEM because it is somewhat straight forward in relation to interaction effects.  But I like the idea that some predictors may have multiplicative interactions and could justify using NPMR.  I work mostly with freshwater aquatic macroinvertebrate and zooplankton so there are many inherent problems associated with these taxa that are always of concern.  Thanks! 

--
You received this message because you are subscribed to the Google Groups "HyperNiche and NPMR" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hyperniche+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[paish...@gmail.com]

It does help indeed.  Looks like I will end up using both but for different studies and objectives.  Thanks Bruce!