Read on Twitter
THREAD: participation bias in behavioural studies during COVID-19 pandemic may be related to trait anxiety, potentially. Please RT!
1/x
This is rather raw, but I feel like it might be important to share this information. In short, participants high in trait anxiety may be less willing to participate in lab behavioural experiments in COVID times, and this is not due to the electrical stimuli involved.
This is rather raw, but I feel like it might be important to share this information. In short, participants high in trait anxiety may be less willing to participate in lab behavioural experiments in COVID times, and this is not due to the electrical stimuli involved.
2/x
We invited 81 participants from an ongoing on-line longitudinal study to take part in a behavioural experiment which would happen in person. This is a well-characterized cohort that we follow since early 2020.
We invited 81 participants from an ongoing on-line longitudinal study to take part in a behavioural experiment which would happen in person. This is a well-characterized cohort that we follow since early 2020.
3/x
In email 1, we mentioned the duration of the session and the reimbursement level (above competitive for Berlin standards). 43 participants expressed willingness to take part and they were sent a second email.
In email 1, we mentioned the duration of the session and the reimbursement level (above competitive for Berlin standards). 43 participants expressed willingness to take part and they were sent a second email.
4/x
In the second email we explained the procedure and the task in detail. Crucially, the study involves electrical stimuli which we thought might be a deterrent to some.
In the second email we explained the procedure and the task in detail. Crucially, the study involves electrical stimuli which we thought might be a deterrent to some.
5/x
Next, we asked whether anxiety (STAI-TRAIT) is associated with willingness to participate. I will say in advance that the two-sample t-test is NOT significant, if p>0.05 is what matters stop reading here, null hypothesis stands! (assumptions likely violated btw)
Next, we asked whether anxiety (STAI-TRAIT) is associated with willingness to participate. I will say in advance that the two-sample t-test is NOT significant, if p>0.05 is what matters stop reading here, null hypothesis stands! (assumptions likely violated btw)
6/x
Distribution of TA scores as a function of the two emails. Raincloud plots.
(Titles are paraphrased)
Distribution of TA scores as a function of the two emails. Raincloud plots.
(Titles are paraphrased)
7/x
Here are the same data using a different plot, histograms and density plots normalized per group. It would appear that quite a few participants with TA 40-55 seem to be rather reluctant.
Here are the same data using a different plot, histograms and density plots normalized per group. It would appear that quite a few participants with TA 40-55 seem to be rather reluctant.
8/x
We donât know why people choose to not participate (few responded to a follow up email).
We also donât know whether there is any meaning to this, t-test says: âNein!â.
And finally, if there is some meaning, we donât know if this is due to COVID.
We donât know why people choose to not participate (few responded to a follow up email).
We also donât know whether there is any meaning to this, t-test says: âNein!â.
And finally, if there is some meaning, we donât know if this is due to COVID.
9/x
All we know is that in our final sample only 5 (out of 28) final participants have trait anxiety higher than 40 (population median). It also appears that most high trait anxious individuals are lost at the first email level, and therefore not due to the electrical stimuli.
All we know is that in our final sample only 5 (out of 28) final participants have trait anxiety higher than 40 (population median). It also appears that most high trait anxious individuals are lost at the first email level, and therefore not due to the electrical stimuli.
10/x
Are our data biased? Does anyone have a similar observation? Comments are very welcome!
(P.S.: If someone can remind of the recent large sample (n=~3500) paper that compared behavioural vs MRI studies as a function of trait anxiety, that would be great, I canât find it)
Are our data biased? Does anyone have a similar observation? Comments are very welcome!
(P.S.: If someone can remind of the recent large sample (n=~3500) paper that compared behavioural vs MRI studies as a function of trait anxiety, that would be great, I canât find it)
