Gavin Simpson
Reputation
15,947
Top tag
Next privilege 20,000 Rep.
Access 'trusted user' tools
 Apr 28 comment Interpreting negative regression intercept; proportion response to categorical predictors There's usually a link function involved, say the logit, so you'd need to apply the inverse of this link function to get things back on the 0,...,1 scale. Sounds like it would be easier to predict from the model for the different combinations (on the response scale) though, from an interpretation point of view. Apr 26 awarded Necromancer Apr 18 revised Find data and confidence “ellipses” (regions?) for a bivariate median? added 524 characters in body Apr 18 comment Find data and confidence “ellipses” (regions?) for a bivariate median? @whuber Ah, I might have misled there. I'll add a new fig that is a real eg of the use case. A dissimilarity matrix computed from the original data is embedded in a Euclidean space using PCoA. But what I neglected to mention is that we compute the spatial medians in this Euclidean space for groups of data points. Hence whilst the x and y are orthogonal over all groups, within an one group there may be correlation. See the updated figure in a minute for an illustration. Apologies for this; I didn't appreciate the importance of certain aspects of the real uses case when I posted the Q. Apr 18 revised Find data and confidence “ellipses” (regions?) for a bivariate median? add note about what median we used, plus that the data are eignevectors of a principal coordinates analysis Apr 18 comment Find data and confidence “ellipses” (regions?) for a bivariate median? +1 Thanks @AndyW - I'd entirely forgotten about the bagplot (guess that's what you get for not teaching my EDA lectures for some years now - totally slipped my mind!) I should have indicated the type of median I had in mind --- I'll update the post, but we've calculated the spatial median, the point that minimises the L1 norm of the distances of the data points to that point. Apr 18 asked Find data and confidence “ellipses” (regions?) for a bivariate median? Apr 14 awarded gam Apr 13 awarded Enlightened Apr 12 awarded Nice Answer Apr 9 comment What does overlap of bootstrap means 95% confidence interval dotplot infer? No, I understand that; my point was, why even bother with trying to interpret overlaps when you are doing the bootstrap anyway? I haven't fully read the paper you linked to but it seems like they are suggesting ways to calibrate reading things off plots other people have produced. If you can avoid this by computing a CI for the thing of interest, that surely has to better? As to the original Q, I doubt these things hold for bootstrap samples, unless the bootstrap distribution is well approximated by a Gaussian distribution. Apr 9 comment What does overlap of bootstrap means 95% confidence interval dotplot infer? For a difference of means wouldn't you compute the observed difference and then the differences for the bootstrap samples and form confidence intervals using these values for the difference itself, rather than for individual means and trying to interpret overlap? Apr 2 answered What does a wedge-like shape of the PCA plot indicate? Apr 1 comment What does a wedge-like shape of the PCA plot indicate? I assume all the variables are positive (or non-negative) & continuous? If so, the edges of the wedge are just the points beyond which the data would become 0/negative. Further, you can get the same pattern you show with positive right-skewed variables; the observations are clumped up at the low end. If you had positive uniform random variables you'd see a (rotated) square. Hence patterns like the one you show are just constraints on the data. Other patterns can show up, like a horseshoe, but these aren't due to constraints on the ranges of the variables. Mar 24 comment R binomial GAM weight Right; it wants integer total counts because that's how the likelihood is defined for the binomial distribution. If you want to get rid of the warning use quasibinomial so that you're just relying on the implied mean variance relationship, and not saying the y is conditionally distributed binomial as well. But to be honest, I don't think you need to account for the imbalance in most cases. Mar 24 comment R binomial GAM weight I think you need to consult some of these Q&As if your question is about imbalance in the the proportions of 0s and 1s. unless you have very skewed proportions, I don't think it makes much difference. You can't supply non-integer weights like this with GAM/GLM, at least not that i have ever seen working. Mar 23 comment R binomial GAM weight I've added an answer but this is more an extended comment. I'm still not clear on what you are trying to do or what your data look like / how they were generated. Perhaps the extended comment will elicit the missing information. Mar 23 comment R binomial GAM weight Can you explain what you are doing? R expects the data to be provided in one of several formats. If you provide response data that is decimal values, you need to also give it the total counts. Mar 23 revised R binomial GAM weight edited title Mar 18 comment Model construction: How to build a meaningful gam model? (generalized additive model) Unless you really need the sparsity-handling capabilities of gamm() or gamm4(), you can use ti() with gam() and use the random effect spline basis bs = "re" to add random effects. itsadug is just some extra functionality built upon gamm() models. If you are fitting Gaussian models, I'd stick with gamm() or even just gam() with random effect splines.