Week 4-Mediation analysis Flashcards
What is mediation?
-A mediator links 2 variables
-The mediator also effects the DV
-So the effect of the IV on the DV (IV-DV) is at least, partially dependent on the mediator (IV-M-DV)
What is mediation often confused with?
Moderation
What is the difference between a mediator and a moderator?
-A mediator is a variable that accounts for an association between a predictor and a DV HOWEVER a moderator affects the strength of a relationship between a predictor and a DV
-A mediator explains why X is associated with Y HOWEVER a moderator will say in what circumstances, associations between X and Y will be apparent
-A mediator must be associated with the IV AND DV HOWEVER a moderator does not have to be associated with the IV and DV (but can be!)
What is a key problem with mediation analysis?
-Another key problem with many mediation analyses in the literature is that they are grossly underpowered
-If you expect a medium strength IV-M and medium strength M-DV association and you were to use the causal steps approach you would need 397 participants (if small effects are expected you would need 21,000 participants!!!!)
-It’s so weak the method it can’t detect mediation is what the small effect point is suggesting
-By contrast the method we look at today would need 74 participants
What is the causal steps approach?
An approach that:
-Has little or no sensitivity at all (needs a huge sample)
-Is mathematically incorrect (IV DV association not necessary)
-Is unable to detect suppression effects
-Mediation analysis is also greatly underpowered
Bad Solutions: What is the Sobel Z Test?
-Gives a p value for the indirect effect
-Based upon a product of the coefficients calculation (i.e. multiplies the regression coefficients together and tests for the significance of the outcome)
-Assumes the product of the coefficient is normally distributed – this is almost never the case
-This method also requires a more participants that to detect indirect effects than the methods described today
What is the Joint Significance Test?
-This method simply ignores the IV-DV association (i.e. it doesn’t have to be significant!)
-If the a-path (IV-M) is significant and the b-path (M-DV) is significant then there is evidence of mediation.
-Joint significance test just looks at a-path and b-path
-However I do not think this is enough- confidence intervals for the indirect effect should be reported too (as well as the indirect effect itself – the B and its SE).
What does mediation analysis test?
-Mediation analyses do not show causal relationships!
-They test proposed associations between variables
How do you write up a Joint Significance Test
According to the joint significance test we have evidence for mediation as there are significant IV-M and M-DV associations, and you can write them up as:
-There was a significant positive association between personality disorder and enhancement motives (B=0.82, SE=0.13, p<.001) with greater personality disorder symptoms being associated with increased enhancement scores. There was also a significant positive association between enhancement motives and unit consumed after controlling for personality disorder (B=0.64, SE=.31, p=.038)
-Shows evidence for mediation
In order to do our additional test for mediation we need to leave SPSS:
-The method we will use calculates 95% confidence intervals for the indirect effect
-https://amplab.shinyapps.io/MEDCI/
What do Confidence Intervals reflect?
-95% confidence intervals reflects how confident we can be in our regression coefficient, it expresses the precision of our estimate, 95% of samples from this population will fall in this range (if we give 95% CI’s we could give 99% CI’s etc)
-High precision= “tighter” CI, this is a good thing it shows consistency in an effect.
-If it overlaps with 0 this means there will not be a significant effect as the range of predicted values overlaps with no effect (0= no change)
-If they don’t overlap you have a significant effect p<.05 (if 95% CI’s)
How would you do the Write-up for this?
There was a significant positive association between personality disorder and enhancement motives (B=0.82, SE=0.13, p<.001) with greater personality disorder symptoms being associated with increased enhancement scores. There was also a significant positive association between enhancement motives and unit consumed after controlling for personality disorder (B=.64, SE=.31, p=.038). 95% confidence intervals (CI) for the indirect effect were computed using RMediation (Tofighi & MacKinnon, 2011), this revealed that there was evidence for a significant indirect effect of personality disorder on drinking via enhancement motives (B = 0.52, SE=0.26, 95%CI .03 to 1.07).
What is Mediation beyond simple effects?
-Mediation analyses discussed today are the first step into more complicated statistical modelling of data
-There has recently been a development of new programs, aided by faster computing, that allow use to look at much more complex relationships between variables
-Much of this involves a technique called bootstrapping
Provide a Summary of Mediation Analysis
Mediation analyses are poorly understood both in terms of:
-Conceptualisation (e.g. confusing it with moderation)
-Analysing (using mathematically incorrect methods)
-Simple mediation can be tested with the joint significance test
-Rmediation can be used to generate confidence intervals for the indirect effect
-More advanced mediation models are becoming more common in the literature, these utilise bootstrapping
-They are relatively easy to interpret (less to run!)
