Psych#Stats

https://www.7bridgeswalk.com.au/fundraisers/sachink/7-bridges-2025

Sachin: walking for a cancer free future I’m taking on 7 Bridges Walk to help those affected by cancer. Please donate to my page. Every step counts because all funds raised will help Cancer Council work towards the vision of a cancer free future. Thank you for your support!

https://www.instagram.com/reel/DLznSXwPajy/?utm_source=ig_web_copy_link&igsh=MXUxNmViaXU0d2x3Yg==

https://www.instagram.com/reel/DLru4TIP5ox/?utm_source=ig_web_button_share_sheet&igsh=Z3Vwc3R2MmFramcx

#psychology #monashuniversity #macquaireuniversity #southerncrossuniversity #wollongonguniversity #unsw #westernsydneyuniversity

Assumption Testing: This is an integral part of any data analysis in student assignments and theses. It's always helpful to have small template handy telling us which assumptions are to be tested for which test. You can be assured of good marks even by briefly covering all the assumptions for the test, and if done thoroughly almost full marks!!
For example in parametric tests there will generally be requirements of normality of something or the other, equality of variance within or between group/groups etc.
I am listing below a few lists which are kind of standard assumptions required for those tests...

Independent samples (2 samples) t-Test : Normality of DV within both groups & Homogeneity of variance between groups

One-Way ANOVA: Normality of DV within each group (levels of IV), homogeneity of variance between groups

Regression: Absence of univariate outliers, independence of errors (easily confirmed by Durbin-Watson statistic), absence of singularity ( no correlations between IVs should be >.9) , absence of multicollinearity ( tolerance should be greater than .1 or conversely its reciprocal VIF should be less than 10), absence of multivariate outliers, absence of influential cases, and then normality, homoscedasticity and linearity of residuals.

MLM: Multilevel Linear Models (PSY863)
Marks in final exam depend on hours studied. Or quality of life (QOL) varies with depression.
In case of marks, what if mark-hours relationship varies with the school attended?
What if the QOL-Depression relationship varies with the clinic that someone attends?
Either, you can do separate analysis for each school/clinic and get lost in murky world of 10-100 regressions OR you can analyze the combined data under MLM!
But what about intra class correlation (ICC), RANDOM intercepts and RANDOM slopes, you might wonder…….
You can think of ICC as the measurement of effect that the school/clinic has on the marks/QOL.
In plain-speak, random intercepts are nothing but the measurement of difference between the schools and then random slopes are measurement of how the effect of hours/ depression changes between schools.
If there is evidence of random slopes, that means the effect of hours/depression is NOT the same between schools!!

Moderation & Mediation Regression
Hi All,
Just trying to put moderation and mediation analysis in plain-speak as I am finding that there is some confusion about these two relatively easy to understand concepts.

Moderation: Lets look at a real life and seemingly logical positive association... Alcohol consumption causes increase in domestic violence ( it could be physical/ psychological or combination of both) - agree?
So we have a positive predictive relationship between Alcohol Consumption (IV) & domestic violence (dependent variable).
Now lets add another predictor - religiosity as a predictor of domestic violence.. What do you think will happen?

.

.

Don't you think highly religious people may be quite averse to domestic violence. And further even if they are intoxicated with alcohol the fear of God in their mind would control the tendency for any kind of violence.

So, there will be a difference between 3 levels of religiosity (high medium &low) on how it affects the relationship between alcohol consumption & domestic violence! There will be an increasing trend at low religiosity, a 'moderately increasing' trend at mid level of religiosity and finally almost a flat line or even decreasing trend at high level of religiosity. THAT is moderation!!! (Figure attached)

Mediation: Let's say high grades in year 12 cause happiness in people. Certainly we will find evidence that there is a positive association between high grades and happiness so we will find a predictive relationship between grades and happiness. Now lets say we throw in another variable that of self esteem or confidence as IV . Suddenly we will find that grades are no more a significant predictor of happiness??? What just happened here?
When we take a step back and check the relationship between grades and confidence, we find that grades actually boost confidence, AKA grades positively predict confidence. Therefore the 'total effect' of grades on happiness is actually made of very little 'direct effect' (the non significant relationship when confidence is added to equation) and relatively large 'indirect effect' through confidence.
What happens here is that grades actually boost confidence which "in turn" increase happiness.

OR

The effect of grades "flows through" or in reality "routed through" confidence to happiness.

I hope the terms in quotation marks - "in turn" / "flows through" /
"routed through" make sense. Because THAT is mediation!!!

Free 1on1 Zoom Session on Moderation Analysis

I have recieved a quite a few enquiries for interpretation of Moderation Analysis using PROCESS. I will be running 6 free of charge zoom sessions of 30 minutes each between 6 to 6.30pm & 9 to 9.30pm starting from Monday 21 September to 23 September. Please message here.

Statistics = Rocket science
Yes / No

No Entry, One-Way, Two-Way etc. etc…
ANOVA
Easiest way to get our heads around ANOVA is to get some kind of picture in our head. Maybe different colors or patterns representing the ever confusing – this way, that way, one way, four way, this effect that effect…
I am attaching a quick reference page attempting to offer you a line sketch of that colorful picture and hope – “You get the picture!”
This is based on fact that essentially ANOVA is a ‘cell design’. This means that we take information on bunch of people who are almost alike in all aspects except their names are different and put all that information in one cell. I just explained the jargon- “Group Mean”  
Anyway, the page builds on this ‘cell’ and shows some elementary ‘Between subjects ANOVA’ & main effects. At the end there is an example of ‘Within Subjects ANOVA’. The purpose is to demonstrate – Why we require less participants in within subjects design?
SPSS, Stata or any other software on which we can run ANOVA, compares these "cells" and tells us how likely its is ( AKA: p value) for all these cells to be nearly equal. If that p value ( or Sig. as in SPSS) is below 0.05, we say it's significant!
Any questions , please contact –
https://www.facebook.com/PsychStats.Au

Psych#Stats Zoom/Skype tutoring services offered for: Statistical Analysis using SPSS/STATA/jamovi/Excel
Subjects: Research Design, Biostatistics, Economics
Unis: ACU, ANU, Charles Strut, Macquarie, Monash, Swinburne, UNSW, U Syd ....