Gimme for continuos data

[Darias Holgado]

Hello all,

I have a question regarding the use of gimme.

I have asked participant to draw how they felt after performing a cognitive task.
I asked for 6 different variables (e.g., mind-wandering, boredom, etc.). Attached one example.

This draw i have transformed into a continuos variable to express the intensity of the sensation across time. Thus, i have "n" data points (depending of the duration of the task, i can transform the draw in more or less data points).

I would like to see the relationship between the different dimension collected and potentially the impact of different conditions.  Looking for solution, i found this "gimme" function, but i am not sure if it is applicable for this idea or this function is more oriented to longitudinal data across time.

In summary, for a specific study, i have 60 participants, with 6 subjective dimensions and each dimension is 60 data points. We want to see the relationship between these dimensions considering the repeated measures at individual level.

Best,
Darías

[Katie Gates]

Hi Darías, 

This seems like interesting data. Thanks for sharing the specific qualities (number of people and time points, number of variables). It seems the data is well-suited for the algorithm. 

Best,

Kaite

[Katie Gates]

Also, you may want to detrend your variables before using gimme. It looks like you have a very clear downward trend there, and gimme (like many time series methods) assumes the data are stationary and do not change in mean over time.

GIMME is looking at how variables covary across time, rather than mean changes. Hope this helps!

[Darias Holgado]

Hello Katie,

thanks a lot for your answer.  Regarding the detrending, do you have any special methods recommendation to do it? 
See attached the overall trend of my data. I have these 6 dimensions, under 3 different stimulation conditions, and the fluctuation of the subjective feeling is not the same for each one.

Best,

Darías

[Katie Gates]

This function would do it: https://www.rdocumentation.org/packages/pracma/versions/1.9.9/topics/detrend 

You'd loop through each variable for each person and save the output from this. 

[Henry Whitfield]

Hi Katie,

Does S-GIMME do any detrending of it's own. I was told it did but perhaps this was phased out?

My data don't have steep inclines like the images in this discussion earlier. Largely the graphs are either horizontal or moving both up and down. However the data where collected during an active psychotherapy which is supposed to influence the measures somehow. Two therapies correlate to a degree with different S-GIMME subgroup models and CS-GIMME models have differences too. Do you have any suggestions regarding how to deal with this stationarity limitation in the write up of such a study?

Many thanks,

Henry