[R-sig-ME] Lack of Variation in Slopes

[Ben Bolker]

On 13-09-17 09:18 AM, AvianResearchDivision wrote:
> Ben,
> 
> What are the chances that there is absolutely no variation in slopes?  This
> seems beyond unlikely, but is probably rooted in my limited understanding
> of what the model is actually doing.  It seems you would always expect some
> sort of variation, even if it is the smallest amount.  Maybe I am
> misinterpreting what this means.  Anyways, thank you very much for taking
> the time to answer my many questions up to this point.  I do very much
> appreciate it.
> 
> Jacob

 It's not that there is no variation; it's that the maximum likelihood
estimate is that there is no variation.  This happens a lot in mixed
models, especially when the variance is small or the data set is small
or noisy.  See e.g. http://rpubs.com/bbolker/4187

 [cc'ing back to r-sig-mixed-models ]


> 
> 
> On Tue, Sep 17, 2013 at 9:06 AM, Ben Bolker <bbolker at gmail.com> wrote:
> 
>> AvianResearchDivision <segerfan83 at ...> writes:
>>
>>>
>>
>>   [snip]
>>
>>> You were spot on with your suggestion.  Thank you very much for your
>> help.
>>> I had mentioned before that I probably should keep random slopes within
>> my
>>> models since that is a primary interest of my study, regardless of
>> whether
>>> a LRT suggests they are needed.  Would you still recommend keeping them
>>> within the model?
>>>
>>> Jacob
>>
>>   I don't think it will hurt.
>>
>> _______________________________________________
>> R-sig-mixed-models at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>>
>