Significant t-test result but non-significant mediational regression result?

#1
Please help! I'm so desperate, this is for my thesis due soon. What does it mean if your t-test is significant but the regression is not?

I am running a psychology study. We assessed two meditation interventions (mindfulness versus loving kindness) and proposed a mediator (self-compassion) and DV (pain tolerance). The study took place across five days (five days of meditation). We predicted that loving kindness meditation would predict increases in self compassion from time 1 to time 5. We predicted that loving kindness meditation would predict increases in pain tolerance from time 1 to time 5. We predicted that self compassion would mediate this relationship between loving kindness meditation and pain tolerance.

When I looked at participants self compassion scores in the loving kindness meditation group, a t-test showed that they were significant from time 1 to time 5. T test also showed that pain tolerance was significant from time 1 to time 5 in the loving kindness condition. It was not significant for mindfulness. However, the mediational model (we used Hayes process model) showed that condition did not predict the increases in self compassion or increases in pain tolerance.

I'm trying to work out; does this mean that loving kindness meditation did increase self compassion? or does it mean that it didn't?

Please help! I would really appreciate it. Thank you!
Lydia
 

hlsmith

Not a robit
#2
Post your output. Ideally, draw out this model and add the values you got from your output. How do you know where "self-compassion" fits into this model? Why do you think it mediates intervention vs intervention mediates it?
 

Karabiner

TS Contributor
#3
I'm trying to work out; does this mean that loving kindness meditation did increase self compassion? or does it mean that it didn't?
In your model, the increase of self compassion in theloving kindness group seemingly was not statistically significantly larger than in the mindfulness group. But we need much more information about the analyses, as hlsmith already mentioned. For example, discussions about studies without knowing the sample sizes are often meaningless.

With kind regards

Karabiner