GIMME Cannot Handle NA Values in Binary Impulse Vector (conv_vars)

[Jessica Birg]

Hi there, 

I’m running gimmeSEM() in R and specifying binary convolution variables (conv_vars) for smoothed Finite Impulse Response (sFIR). However, I keep encountering the following error:

gimme ERROR: missing values in the binary impulse vector not allowed

The issue arises because my dataset contains NA values in conv_vars. Specifically:

• My binary impulse variables are socint (social interaction: 0 = no, 1 = yes) and restype (prompt answered: 0 = skipped, 1 = answered).
• If restype = 0, the participant did not answer the prompt, which means socint is NA for that time point.
• However, GIMME does not accept missing values in conv_vars, and I cannot assume socint = 0 because a skipped prompt ≠ "No interaction."

I'm wondering if I should recode the NA values as something else or just remove the missing prompt rows altogether. I wanted to use the restype variable as exogenous to see if there was a pattern to the missingness. Curious what others have done to get around this issue. Thanks!

Jessica

[Jessica Birg]

I’m running gimmeSEM() in R and specifying binary convolution variables (conv_vars) for smoothed Finite Impulse Response (sFIR). However, I keep encountering the following error:

sql

CopyEdit

gimme ERROR: missing values in the binary impulse vector not allowed

[Katie Gates]

Hi Jessica, 

The work around for sFIR is to impute values. So, you would have to figure out what would be a reasonable value to impute there. I don't know of another solution. Here is some documentation on the function we use: https://www.rdocumentation.org/packages/gimme/versions/0.7-18/topics/sFIR

Best,

Katie

[Jessica Birg]

Hi there, 

I was able to finally run my gimme model on two datasets after much trial and error and was wondering if anyone had some time to help me understand/dissect the results accurately? 

On the first dataset, I ran a group gimme (subgroup = false) and then tested whether beta coeffs differed in strength by SIAS (social anxiety) score. In the second dataset, I ran CS-gimme, specifying two subgroups (0=healthy controll, 1=SAD) and tested whether specific subgroup paths differed in strength by diagnosis. I'm feeling a bit overwhelmed with the number of output files and was hoping someone could help me break down what I'm seeing here at a basic level so I know I'm interpreting results accurately. 

Does anyone have R code they would be willing to share with me that visually depicts different path strengths by a third variable (SIAS or SAD dx)? Or any helpful visualization code would be appreciated in aiding my interpretation of results! 

Thanks!!

Jessica

--
You received this message because you are subscribed to the Google Groups "gimme-r" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gimme-r+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/gimme-r/dc69a5b4-a416-496b-aa6b-a73f7a845bf0n%40googlegroups.com.

[Dina Dajani]

Hi Katie,

I had a similar question -- 

I am using as input to gimme task data that is concatenated across Runs 1,2,3 and 4. The subject matrices includes time courses for relevant ROIs and binary vectors representing two different task trials. Before concatenation, I added a row of NAs to the bottom of Runs 1,2 and 3 matrices (including for the binary task trial columns) to indicate discontinuity between runs. But, that resulted in the same error as below "missing values in the binary impulse vector not allowed". Should I replace the NAs for the binary task trial vectors with 0s instead? Thanks

Dina

[Katie Gates]

Hey Dina, 

Yes I'd replace the trial vectors with 0s. That should fix it without any unintended consequences for the HRF estimates. 

Katie