Subscribe: Comments for i'm a chordata! urochordata!
http://www.imachordata.com/?feed=comments-rss2
Preview: Comments for i'm a chordata! urochordata!

Comments for i'm a chordata! urochordata!



a blog about marine biology, ecology, and a little helping of statistics



Last Build Date: Mon, 09 Apr 2018 17:18:20 +0000

 



Comment on Do Not Log-Transform Count Data, Bitches! by jebyrnes

Mon, 09 Apr 2018 17:18:20 +0000

Are your residuals non-normal, or is it your data? If your data, meh. It's all about the action in the residuals. If they are non-normal, and this data is not bounded by 0, you might want to think about either finding a different error distribution to use for your model or, if the problem is heteroskedasticity, using a generalized least squares approach where residual variance is allowed to vary by treatment. Does that make sense?



Comment on Do Not Log-Transform Count Data, Bitches! by Yair

Wed, 28 Mar 2018 18:10:23 +0000

Hi, I found this article very interesting and I´m writing this hoping you could shed some light on an analysis I´m performing. I am trying to analyze some data about animal behaviour and would need some help or advice regarding which non-parametric test should I use. The variables I have are: -Response variable: a continuous one (with both positive and negative values) -Explicatory variable: a factor with 6 levels -Random effect variable: as the same animal performing some behavioural task was measured more than once. As I have a random effect variable, I chose a GLM model. Then, when checking the normality and homoscedasticity assumptions, Shapiro-Wilks test showed there was no normality and QQplots revealed there weren´t patterns nor outliers in my data. So the question would be: which non-parametric test would be optimal in this case, knowing that I would like to perform certain a posteriori comparisons (and not all-against-all comparisons)? My database has lots of zeros responses in some conditions, I´ve read that for t-students tests lacking of normality due to lots of zeros it´s OK to turn a blind eye on lack of normality (Srivastava, 1958; Sullivan & D'agostino, 1992) ... is there something similar with GLM? Thank you so much in advance for any advice you could provide. Kind regards, Yair Barnatan Ph.D. Student - Physiology and Molecular Biology Department Faculty of Science University of Buenos Aires



Comment on Do Not Log-Transform Count Data, Bitches! by jebyrnes

Wed, 20 Sep 2017 20:09:52 +0000

Well, if you have negative numbers, then you've got a different distribution on your hands! So, often, a rethink is needed. I've lately become a fan of using weighting to accommodate for different variance structure in normal distributions, but there are a variety of options - see https://biol609.github.io/lectures/04_gls.html#/section for some options and examples.



Comment on Ecological SEMs and Composite Variables: What, Why, and How by jebyrnes

Wed, 20 Sep 2017 20:05:50 +0000

Also, sorry for the gap - was in the field!



Comment on Ecological SEMs and Composite Variables: What, Why, and How by jebyrnes

Wed, 20 Sep 2017 20:05:34 +0000

Yes - this post was written before lavaan added a specification for composite variables. It actually did so because of conversations I had with Yves, and I think internally uses the specification I put above to resolve them. They're still tricky, but much easier to use now than they used to be!



Comment on Ecological SEMs and Composite Variables: What, Why, and How by Tom

Fri, 11 Aug 2017 15:02:33 +0000

Sorry, comments need to be removed for the code to run.



Comment on Ecological SEMs and Composite Variables: What, Why, and How by Tom

Fri, 11 Aug 2017 14:49:09 +0000

Hi Jarret, Thanks for this interesting post. I've been playing with the Cardinale composite example above and seem to be obtaining the same answer with this simpler model specification: compositeModel2<-' Nitrogen <~ logN + 1*logNcen2 # change '~' to '<~' for the formative construct Nitrogen ~~ 0*Nitrogen # Specify 0 error variance # Nitrogen =~ SA # SA ~~ SA SA ~ SR + Nitrogen # specify regression on SR and Nitrogen composite logNcen2 ~ logN ' compositeFit2 <-sem(compositeModel2, data=cards) # no need for 'std.lv=T' Do you agree they are the same, and would this mean the trick in the blog post for specifying the composite is no longer required? Thanks!



Comment on Do Not Log-Transform Count Data, Bitches! by nadiab

Fri, 28 Jul 2017 09:26:48 +0000

Great, finally someone that speaks about statistics in an actual comprehensible way!! I am an eternal beginner in stats (meaning that I can learn things and do them right but then I forget everything and have to start from zero) but this I can understand. Thank you so much!



Comment on Introducing the Open Derby (guest post) by Reproducible ecology with ‘Open Derby’ – James Robinson

Wed, 05 Apr 2017 02:22:01 +0000

[…] I’m troubleshooting and exploring how to run Open Derby with amazing Baum lab graduate guinea pigs at the University of Victoria (BC, Canada), while developing the idea for more general consumption with incredible leadership training from Mozilla’s science lab. You can read about the progress of our first Derby on Dr. Jarrett Byrnes’ ‘iamchordata’ blog. […]



Comment on Space and SEMs by jebyrnes

Tue, 04 Oct 2016 13:05:04 +0000

It should be sufficient just to correct the standard errors, as lavaan will still return an unbiased estimator of coefficients. That said, I need to put a warning on this post. Lavaan currently does not produce correct residuals, which means the output of the function is not good. You have to calculate residuals by hand. I think Yves is working on it, though.