r - Weighting the inverse of the variance in linear mixed model -
I have a linear composite model that runs 50 different times in a loop.
Every time the model is run, I want to get the reaction variable b weighted with variation. So if the deviation of B is small then I want to make weighting bigger and vice versa. This is a simplified version of the model:
models & lt; - lme (b ~ type, random = ~ 1 | repeat, weight = ~ i (1 / b)) Repeat:
I have weight options in RLME Trying to do this. Now I have this:
weight = ~ i (1 / b) But I do not think it's right ... maybe weight = ~ 1 (1 / var (b)) ??
I also want to adjust it a bit because there are two types of data specified in the type of variable variables (2 levels) in B.
I want to weight the variance of each of these two levels separately. How can I do this?
I'm not sure that it is understandable to talk about loads feedback < / Em> variables in this manner The description I have found in the mailing list of R-SIG-Mixed-Models refers to the use of inverse weight from the predictor variable, either fixed effects or random effects. Fits to use load feedback in reducing the deviation of the model's approximation. There is a function that gives definite impact variation (subclass of varFunc function family) and there is a help page (? Gls page): ? VarFixed? VarFunc is required for its logic as its formula. So my original guess was:
model which you had proven wrong. If this works, then see:
Model & lt; (I have a constant estimate that this weight is the default position and these special loads There is no need to specify that the inversion of the load is implied and there is no need to be said explicitly with "1 / type". In the case of mixed models, "correct" construction depends on the design and pre-science. And none of these presented So it is actually only a syntactic comment and this model is not supported. I have not downloaded the files It seems that there are three different files and there is no code to add them to the dataframe. In the same way, the same would be a data object, within which the column name will be used in the sources of the regression function (I also suspect that this is the default of this function Behavior and therefore my unintended prediction is that you will not make any changes to the OM by opening this 'weight' parameter.)
Comments
Post a Comment