Dear, everyone
I work with longitudinal data for growth model using nonlinear mixed model basing on logistic growth equation. I already fitted the model, but I am confused about what should be reported and especially plotted in results. Can anyone give me some advices or recommendations? Thank a lot.
P/s: I use R in my work.
I have only started learning this myself. This is a big topic, whole chapters have been written on it. I find the book "Applied Longitudinal Data Analysis" by J. Singer and J. Willett very useful, so far. It is very approachable, very intuitive writing style, and there is a chapter (ch. 6) on modeling discontinuous and non-linear change with examples on how to plot time trajectories (but no R code). There is another book on Longitudilan modeling using R, but I have only just started reading it J.D. Long "Longitudinal Data Analysis for the Behavioral Sciences using R". This, too has a chapter (ch 7) on modeling non-linear change and examples on how to plot growth curves and polynomials in R. It also has examples on R code for that. So, this might be really useful to you, since you are using R. I find it very well explained, so far.
It depends on what methods you used to fit the model.
In any case, I think the analysis should include tests that you used to validate your model.
Also, the confidence intervals for the parameters is a very important part of estimation results.
Thank you @Stefanie Doebler and Andrey Davydenko for your useful suggestions.
This is meant to be helpful! The real thing you need is material that conveys the evidence to the research questions you have asked, So work out precise questions first and then answer those,
Thank you @Kelvyn Jones for your advices. I think you are familiar to nonlinear mixed model with R. In my work , I want to examine the relationship of some covariates on the fixed effects of parameters ( 3 parameters in logistics growth curves). Interestingly, I found that there is interaction of 2 continuous covariates on 1 parameter. However, it is also challenge for me to explain the interaction effects by using graphics. Can you give some suggestions? Thank a lot.
You need to create a prediction data set
1) take some characteristic values of the two variables forming the interaction say deciles so that would be 100 (10 by 10) values . Use these values and their interaction to predict
2) For all predictors other than those involved in the interaction set them at their average value - so you will have a 100 of these but with all values constant for each predictor.
3 Use these values and the estimated fixed coefficient to make predictions say in a spreadsheet
4 Plot the predicted values against one of the X variables involved in the interaction and plot a separate (group) line for the deciles of the other predictor involved in the interaction - it is worth playing around which variable is grouped and which one is on the X axis
MLwin does such mixed model and has customized predictions facility to do such predictions and plotting
https://www.researchgate.net/publication/234015221_Manual_supplement_for_MLwiN_Version_2.14
Data Manual supplement for MLwiN Version 2.14
Thank you so much @Kelvyn Jones for your kind supports. I will try to get it ^ ^
- Badges
- Science topic
- Similar topics
- Mathematics
- Algebra
- Matrix