18.20 Concluding remarks
By rmjlevb, on 12 May 2014
Thank you for those who attended today and those of you who have been following online. It has been as stimulating message. This session has focused on applying Bayes factor to an example of a smoking cessation intervention. Zoltan has a calculator on his website you can use. However, your data needs to be normal. For those, of you who want to do more complex analyses with different distributions I recommend Winbugs.
18.16 Finishing off with questions and potential applications
By rmjlevb, on 12 May 2014
Is robust to violations of assumptions? More robust than traditional tests. But still sensitive.
18.09 What about referencing and funding?
By rmjlevb, on 12 May 2014
Will you get published if you use it? Bayes has been talked about for decades but is now taking route. Publication is getting easier.
18.01 Analysis using Bayes factor
By rmjlevb, on 12 May 2014
Recommend if you are interested: Dienes Z Understanding Psychology as a Science. London: McMillan 2008 – A superb and very clear exposition of philosophy of science, and classical and Bayesian statistics
17.51 Continued . . .
By rmjlevb, on 12 May 2014
Robert and Zoltan are going to take us through the process:
Primary outcome: attitudes
– Attitudes score using surveys which display various positive and negative views about smoking.
-Pre-test score was 4 (ranged from 0 to 10) for girls
-Pre-test score was 5 for boys
-Post test one year later
Analysis
Use the change score – post and pre-test score to see if there has been an effect.
Decisions (forming the prior)
– We would expect in the null case for girls if nothing was done that the score would increase from 4 to 8. So the change would be 4.
– if it had an effect we would expect it to stay the same at 4. So change would be 0.
– Use a half normal distribution with 4 as the point estimate
17.47 Robert West session
By rmjlevb, on 12 May 2014
The major issue with this intervention is that there is no control group. This is a common problem in epidemiology and public interventions. One of the possibilities is to look at Boys. They were not included in the intervention but were exposed.
Another striking thing is that we are looking for no change. We don’t want smoking rates to increase. If prevalence goes up we can assume the intervention wasn’t that really effective.
How are we going to come up with our values for the Bayes factor. What is the shape of the distribution etc?
17.45 Schools programs
By rmjlevb, on 12 May 2014
Try to change the way school sessions are usually run. Being delivered by roll model, out of the class room and focus on “gross factor”.
17.44 Seeding strategy
By rmjlevb, on 12 May 2014
Included story lines in popular programs and are getting celebrities to promote awareness of the impact of smoking.
17.40 The teen movement
By rmjlevb, on 12 May 2014
The teen movement is called Sky. For the teenagers the social world is everything to them, trying to find their identity, believe only friends know then, and only thought about here and now. The idea was to work within this social world and social currencies.
The movement is all about being true to yourself “being good without it” is the motto used. They put the choice not to smoke at the centre. The key components of the movement – magazines, pop up events, online information, celebrities etc.
They are currently trying to build the sky movement – you have to take a pledge to be true to yourself. Smoking is embedded throughout the movement.
17.37 continued
By rmjlevb, on 12 May 2014
Saw parents smoke quite often. Unlike here where parents try to hide it. Often family members send them out to try it.
They designed a multi-intervention approach:
1) Teen movement
2) Parents campaign
3) Social sessions
4) Seeding influences, content