Missing data handling & plot path significance

[Kylie F]

Good afternoon,

 

I understand that the GIMME package can handle missing data (by making sure the missing data points are reported as NA), but I wanted to know how exactly the package is handling those missing points. Are they replaced with values directly preceding them, or some kind of average, or omitted?

 

If a path is represented in a plot, does that mean the relationship between those two variables is significant?

 

Thank you in advance for your time!

[Katie Gates]

Hey there, 

Great question. GIMME uses Full Information Maximum Likelihood as provided within the package lavaan for estimation. This provides case-wise maximul likelihood estimation using all information available. Thus no imputing is done. They are omitted. 

Regarding the plots - this depends on which type of plot you are looking at. 

For the group -level plots (with black and grey lines) paths indicate the presence of that path for at least some individuals. Black paths indicate that everyone had the path estimated, and it was significant for at least the proportion of people specified in the "groupcutoff" argument (the default is .75). Grey paths indicate that fewer than this proportion had the path. Line width here corresponds with the percentage of people who have that path estimated. 

When looking at the individual-level plots, a path indicates significance, as long as it's not a group-level path. Since group-level paths are estimated for everyone, sometimes a small proportion of individuals will not have it be significant but it will still be estimated (and present in plots). Heat map convention is used for coloring, with red = hot/high and blue = cool/negative. 

Significance at the individual level for each path can be examined looking at the betas and standard errors for each person. These are in different places depending on if you stored your data in a folder or an object. 

Hope this helps,

Katie