I would really appreciate if anyone could help me with an answer to my problem.
First I will describe my stats.
My sample size is 1822 participants.
I used many questionnaires.
The one that is giving me a headache is a 10 item scale that is in turn a sub-scale of a 20 items scale. The complete scale has a 0.8x cronbach alpha in the review papers. So i though that if i used the sub-scale in my study.... it would also have a good cronbach alpha.... it turn out not to be like that and it has a .484 cronbach alpha... If i delete 1 item it gets as high as .50 but thats about it.
The thing is that in one of my linear regression analysis one of my groups showed a relationship with this scale and the other one didnt. which is GREAT, it is exactly what i was expecting to find...
but then a t-test between groups showed a p = 0.081. which i argue that it could have been significant if i had used a better scale.
but my problem comes when i look at the .48 cronbach alpha, i dont know if i can argue that there is a difference.... or that that regression is actually telling me what i think it is....
This is paper worth 40% of my whole year >_< and my supervisor has a close eye on how i go about this low cronbach alpha and what it actually means.
I've read around the area but I have NO clue when i add everything up on what can or cant conclude because of this 0.48 cronbach alpha, because i dont really know how bad it is given that is a subscale.
I know that .70+ is ideal, 0.60 is ok, and anything less than 0.60 is not great. but how bad is 0.484 or if i decide to take 1 question down to increase it to 0.50 (which i dont think is that much of a difference but might just look good for the reader given that it starts at least w .5 instead of .4)
So my questions are
1) How bad is 0.484 cronbach alpha?
2) Should i take 1 question so it goes to 0.50?
3) What can i really infer from the results when i take the regression, t-test and cronbach alpha together?
any clue or hint or answer or anything is REALLY appreciated.
Thnk you guys