MEA in mother-father-infant interactions

[hervetissot]

Dear Fabian,

I am a new user of MEA. First of all, thanks for putting out such amazing tools. I am conducting research at the CHUV in which I typically use observation of mother-father-infant interactions. I ususally conduct macroanalysis assessments on these data, but am looking for coding my data using your software. I now have a couple of questions:

1) what “thresholds” would you use in the MEAccf function for time-related settings, i.e. lagsec, winsec, and incsec considering the nature of parent-infant interactions (probably shorter “windows” than in adult-adult interactions)?

2) Did you already experience the analysis of triadic interactions? Do you by chance have an example of the procedure you used?

By the way, your 2020 paper about the rMEA package is very clear and “user-friendly”, thanks for that!

Best regards,

Hervé

[fabian]

Dear Hervé,
Thanks for your message!
1) The best settings depend on your video-material.
For parent-infant interaction, I would suggest to not go higher than the established settings for psychotherapy (lag 5s; winsec 60s; incsec 60s):

lagsec=5 (you can check the setting of lagsec in the lagplot of rMEA: where real interactions and pseudo-interactions cross each other, this should be the maximum lag)

winsec=30 (how long are your sessions?) 30 or lower is probably better, as the turns of parent vs. infant are most probably much shorter than in psychotherapy

incsec=10 I would use overlapping windows, especially if you use winsec of 30 or lower.

2) Yes – we analyzed marital therapy. A simple way is to just use all possible combinations.

Cheers, Fabian

[hervetissot]

1) the sequences I started to analyzed are short clips of 1 minute 30 seconds. For the moment, I chose lagsec = 5s and winsec = 10s, as turns in parent-infant interactions are known to happen in shorter windows because infants need greater contingency in interactions (I found study results in a paper by Ed Tronick to support this decision).
As the clips are quite short, I chose short incsec of 1s.

2) OK thanks!

[fabian]

lagsec = 5s and winsec = 10s
–> this may work, but you have to be aware that in such short windows, a lag of up to 5seconds is high (you are losing data at the beginning and end of your short window).
I suggest to lower the lag to 3s or even 2s
Just make sure to carefully inspect your results.

[hervetissot]

When conducting analyses on multiple groups, would you recommend conducting the smoothing and rescaling operations on the total sample or on separate groups?

[fabian]

I would do it on the total sample.
Do you expect large differences between separate groups?

A good way to reach an empirically-based solution is to carefully look at the lagplot (and/or the heatmaps).
These graphical displays should give you a good insight into the decision on separate or group-level pre-processing.
Cheers, Fabian

[Author]

Posts