[R-meta] Accounting for sources of variation in meta-analysis

[Carla Gomez Creutzberg]

Dear Michael,

Many thanks for your reply. Indeed I did not mention that in our current
analysis I have taken log response ratios as measures of effect size. Per
study, I calculate these over all pairwise combinations of land uses based
on their ecosystem service provision values. For this, I first combine
different measures in each study to get an indicator of service provision
per land use.  With the ratios across pairs of land uses for all studies I
can then combine estimate using a network meta-analysis.
With the subset of studies that have biodiversity I would also be able to
derive, per study (and per land use within a study), a single biodiversity
indicator like we did for the ecosystem service indicators. However, I am
not really sure whether and how I could introduce biodiversity as moderator
into the network meta-analysis, which is why I've started looking for
alternative approaches.

Thanks for your attention and your suggestions - I look forward to hearing
any further thoughts you (or any of the other members) may have on this
tricky analysis!

With best wishes,

Carla
On Mar 19, 2018 10:18 PM, "Michael Dewey" <lists at dewey.myzen.co.uk> wrote:

> Dear Carla
>
> Can you expand a little on this? At the moment what estimates go into your
> current meta-analysis? Are they some relationship between land use and
> service provision? And do you have a single potential measure of
> biodiversity for each current estimate? If you do then a first step might
> be to use biodiversity as a moderator in a meta-regression.
>
> Michael
>
> On 19/03/2018 03:53, Carla Gomez Creutzberg wrote:
>
>> Dear fellow meta-analysts:
>>
>> I hope this message finds you all well.
>>
>> I have collected information on the provision of ecosystem services by
>> different land uses and have conducted a meta- analysis on this. However,
>> I
>> now wish to expand this analysis to a subset of our meta-dataset that,
>> besides the effect of different land uses on ecosystem service provision,
>> also has data on the biodiversity of each land use. I would like to use
>> this to explore if biodiversity accounts for some (or all) of the effect
>> we
>> see of land use on ecosystem service provision.
>>
>> After reading a bit, I am beginning to think that one of the ways that we
>> could probably use to approach this analysis would be a long the lines of
>> MASEM (Meta-analytic structural equation modelling) or a structural
>> equation modelling -based meta-analysis. I am entirely sure yet which one
>> applies best to our case yet but probably its the second one, which seems
>> to be not so well documented. In any case, at this stage, both options
>> seem
>> a bit intimidating and complex so I was wondering if anybody knew of any
>> alternative methods to separate the effect of land use and biodiversity in
>> the provision of ecosystem services within our datset? I am not sure if
>> taking biodiversity as a moderator would be of any help with this? In our
>> dataset, for each study, we have information on 2 or more land uses, and
>> for each land use we have an indicator of ecosystem service provision and
>> an indicator biodiversity.
>>
>> What ultimately interests us is the effect of land use on ecosystem
>> services with and without accounting for biodiversity. However, I am
>> concerned that to tease this apart we may need to eventually have
>> biodiversity as a response to land use and ecosystem services as a
>> response
>> to land use as well which could give rise to problems with
>> non-independence
>> since we would have 2 effect sizes per study.
>>
>> Any insights or suggestions on this are certainly most welcome.
>>
>>
>> Thanks for you attention!
>>
>>
>>
> --
> Michael
> http://www.dewey.myzen.co.uk/home.html
>

	[[alternative HTML version deleted]]