Random variable in mixed-effect model (ecological studie) I'm beginner with the mixed-effects model, so I already apologize if my question is a bit naive.
My problem is the following : I sample each time 30 plants in 6 populations on 9 mountains. So I thought that I should write the random effect as (R-package lmer4) : 1|mountain/population (each plant being nested in a population which in turn is nested in a mountain). One of the aim of my study is to test the effect of altitude on sexual reproduction, so a colleague told me that if I put mountain/population as random effect, the model will "level" the differences between populations (exactly the opposite of what I want) and that I should better write : (1|mountain + 1|population). 
Can anyone enlighten me on this point ?
More generally, how can you know if a factor has an effect on the slope or on the intercept ? Is it a visual analysis that permits to decide ?
Thank you in advance,
CL 
 A: Unless your populations are labeled the same on every mountain (e.g. you have populations 1, 2, 3 on mountain A; populations 1, 2, 3 on mountain B; populations 1, 2, 3 on mountain C ... etc.), (1|population)+(1|mountain) and (1|mountain/population) will be equivalent (I would have a mild preference for the latter since it reminds you that the experimental design is nested).
Even if (maybe especially if) altitude is your focal variable, and the among-population differences are just a nuisance, you really need to leave altitude in the model. Since there might be differences in the effect of altitudes among mountains, you probably want something like
sex_reprod ~ altitude + (altitude|mountain) + (1|mountain:population)

... this allows for variation in slope and intercept by mountain and in intercept by population (recognizing that you can't distinguish an among-population variation in the effects of altitude because each population occurs at a single altitude).
Following Murtaugh (2007, Ecology) Simplicity and complexity in ecological data analysis, if you have a reasonably balanced design you might simplify your problem a bit by aggregating the data to the level of the population. (This won't work if you want to fit a GLMM, e.g. if reproductive strategy or status is a binary variable, or if you have individual-level covariates (size, aspect, etc.) in your data set.)
