[fabian]

The issue is fixed in version 5.04:

Fixed the output index for analyses of existing video files. The previous “t_ms” column reflected elapsed processing time rather than video time. It has been replaced by “frames”, indicating the number of processed analysis frames. Existing movement values from version 5.03 and earlier remain valid.

For AppStore users: Please update your version of MEA5
For Windows users: I am working on an easy way to update MEA5 for Windows. For the time being, please ignore the variable “t_ms”.

[fabian]

Hi salamander2 🙂

Thanks for reporting this important issue!

I have tested MEA5 on macOS 14, 15, and 26 and have not been able to reproduce the problem so far. It is very helpful that you used my test video, because this allows us to largely rule out a video-codec-specific issue. The Apple Silicon generation should also not be relevant, as I tested MEA5 on M2 and M4 systems.

One possible explanation could be App Store sandboxing, although I am not yet sure whether this can actually cause the behaviour you observed.

I suggest that you send me an email at mea@psync.ch. I can then provide you with a direct download link to a version outside the App Store, so that we can test whether the issue is related to the App Store build. In parallel, I will continue looking into other possible causes.

Cheers from Bern, Fabian

[fabian]

Hi,

Regarding the sampRate parameter: My understanding is that this should match the frame rate of the analyzed video files (in my case, 25 FPS). Is this correct?

yes – insert the sampRate of your video. This needs to match your video. otherwise you will get wrong timinig information.

The other papers just based their analyses on videos with different frames per second (fps). –> just use the fps-settings of your video

cheers, Fabian

[fabian]

Hi HILA,

I expect around 4,500 lines. However, currently, the output file contains only about 300 lines.

–> Yes, this is correct: Should be 4.5k lines.

I suggest to downscale the video to e.g. 960×540 (for HD) or 640×480 (for 4:3 resolution).

AND: Use a common codec such as h264

Cheers, Fabian

[fabian]

Hi – please check out the instructions and documentation on OSF:
See pages 11-13 on the pdf for MEA:
https://osf.io/gkzs3/files/37szn

Best, Fabian

[fabian]

Hi romi,

as replied in the other forum, I assume that this is a windows-specific problem.
Please try installing Max/Msp from cycling74.com

Best, Fabian

[fabian]

Hi romi,
It could be that you need a version of Max/Msp on your computer:
Specifically, the Windows-version of MEA sometimes needs an installation of Max/Msp from Cycling74, because some necessary components may be missing on your PC.

In other words:
a) please let me know which OS you are on
b) try installing Max/Msp from cycling74.com
–> you do not need to buy a license in order to run MEA

Cheers, Fabian

[fabian]

Hi Pia,
Yes you can easily experiment with other winSec and incSec parameters.
In my experience, the winSec has a much bigger effect than the incSec, so I suggest starting with changing the size of your windows. However, keep in mind that the smaller you choose your winSec to be, the less datapoints will remain for your cross-correlations. In other words: Small windows will also imply short lagSec.

For the lagSec parameter, an empirically based way to determine this value, I suggest to look at the lagplot: The boundaries where your real synchrony crosses your pseudosynchrony may be taken as an approximation for a good lagSec parameter.

Best, Fabian

[fabian]

Hi gr220,

“w” is the abbreviation for “window”, i.e. w1=window1; w2=window2
w2 –2.56 may be interpreted as follows:
In your window number “2”, the lag with the highest cross-correlation-coefficient was “–2.56 seconds”
bestLag thus means the time-delay (lag) where the association between the two time-series was at it’s highest value.

grand average is the mean ccf for all lags and all windows. It may be described as the “overall synchrony” in this specific interaction.

You can find a fully detailed description of all parameters in the following publication:
Kleinbub, J. R., & Ramseyer, F. T. (2021). rMEA: An R package to assess nonverbal synchronization in Motion Energy Analysis time-series. Psychotherapy Research, 31(6), 817-830. https://doi.org/10.1080/10503307.2020.1844334

Cheers, Fabian

[fabian]

Hi Elif Ozan,

Thank you for your interest!
To download, just navigate to “downloads” in the menu, and there simply fill out the necessary information.
Yo will receive an automated answer with links to the download.

That’s it 🙂

Cheers, Fabian

[fabian]

Dear minisharmaa,
You are correct in what you state above. The most straightforward solution to the problem of different sized ROIs is using rMEA (R-package), where you can apply the z-transformation of your raw MEA-data (withoug centering the data).
e.g. –>
## rescale with standard deviation
mea_scaled = MEAscale(mea_raw, scale = “sd”, center = F, removeNA = TRUE)

For more details, see:
Kleinbub, J. R., & Ramseyer, F. T. (2021). rMEA: An R package to assess nonverbal synchronization in motion energy analysis time-series. Psychotherapy Research, 31(6), 817-830. doi:10.1080/10503307.2020.1844334

Best, Fabian

[fabian]

Hi,
Once you save your raw-data (as a .txt-file in the MEA-application), you will have full control over your data.
I suggest using the R-package rMEA: This will allow you to check and analyze your raw-data from MEA in a convenient way.
https://cran.r-project.org/web/packages/rMEA/index.html

Cheers, Fabian

[fabian]

Hi minisharmaa,

I am happy to hear that the analysis worked so far 🙂
Regarding your question:
Is it okay to have a 100% relative amount of synchrony and what does it mean? Please help to understand this lag plot graph and also how I can interpret it?

The lagplot provides both a graphical as well as a numerical estimation of when and for which duration your original data is different from randomness. If your original data did not contain any signs of substantial synchrony, the real data would be similar or very close to the random data (pseudosynchrony).
Your finding of 100% above random means that for the chosen lagsize (lagsec=5), the real data is always higher than the pseudodata. This is “normal” and means that for this extension of time ±5sec, real synchrony goes beyond pseudosynchrony.
I suggest you try out the following: If you increase lagsec (e.g. to 10ec), you will probably see that your real data will cross the pseudodata at some point in time. The crossing-point may be regarded as the limit of the social present.

You can find more information in the paper on rMEA:
Kleinbub, J. R., & Ramseyer, F. T. (2021). Rmea: An r package to assess nonverbal synchronization in motion energy analysis time-series. Psychotherapy Research, 31(6), 817-830. doi:10.1080/10503307.2020.1844334
And in the paper on the social present:
Tschacher, W., Ramseyer, F., & Koole, S. L. (2018). Sharing the now in the social present: Duration of nonverbal synchrony is linked with personality. Journal of Personality, 86(2), 129-138. doi:10.1111/jopy.12298

Good luck and cheers from Bern,
Fabian

[fabian]

You are right: The qualitative aspect of movement cannot be measured with MEA (see: Ramseyer, F. T. (2020). Motion Energy Analysis (MEA). A Primer on the Assessment of Motion From Video. Journal of Counseling Psychology, 67(4), 536-549. https://doi.org/10.1037/cou0000407)

For the triadic aspect, these papers may also help:
Dale, R., Bryant, G. A., Manson, J. H., & Gervais, M. M. (2020). Body Synchrony in Triadic Interaction. Royal Society open science, 7(9), 200095. https://doi.org/10.1098/rsos.200095
Fujiwara, K. (2016). Triadic Synchrony: Application of Multiple Wavelet Coherence to a Small Group Conversation. AM, 07(14), 1477-1483. https://doi.org/10.4236/am.2016.714126

The other aspects (speed, duration, max, etc.) are on the individual level and they can be derived from the time-series generated by MEA. (this was done in an older script in SAS, but I suggest to implement it in R according to your requirements).

[fabian]

Hi,

Yes – this is quite straightforward if you use rMEA
(https://cran.r-project.org/web/packages/rMEA/index.html)
The MEAccf function provides you with a value for overall synchrony (all_lags) and values for either subject1_leading (s1Name_leading) or subject2_leading (s2Name_leading).

If you want to calculate synchrony for more than 2 subjects, you need a different approach, or you could calculate synchrony for all possible pairs.
Best, Fabian

[Author]

Posts