Beyond ANOVA (Part 6.4) {MANOVA & MANCOVA}

[Neeraj Kaushik]

Attached is the file band.sav
Here we've 4 var(s)
Sales, web & material are metric var
Perfo is non-metric {Though no coding is given for it but Use Analyze->Descriptive Statistics->Frequency to see that infact its Non-metric var with 3 categories.
 
We want to check the following Obj:
Obj 1: To check whether web & material together depends on performance.
 
Dep var: web & material
Indep var: perfo
Technique used: One way MANOVA
 
Go to Analyze->General Linear Model->Multivariate
Put web & metrial in dep var
& perfo in Fixed factors.
Click on Options->Homogenity test
Click on Continue->OK
 
Output Analysis:
Table-2 (Box's test) & Table-4 (Levene's test) confirms that assumptions are not violated.
Table-3 indicates perfo is related with dep var(s)
Table-5 indicates perfo is related with both dep var(s)
 
So now we'll repeat the analysis with Descriptives, Post hoc & Plots.
Go to Analyze->General Linear Model->Multivariate
Click on Options->Uncheck Homogenity test (it has been tested)
On top, send perfo to Right hand side (RHS) box of "Display mean for"
Click on Continue
 
Click on Post hoc->send pefo to RHS block and choose method Tukey
Click on Continue then OK
 
Output Analysis:
Look table: Estimated marginal means
For both dep var(S) mean scores for 4 are less than 5 & 6, while mean scores of 5 & 6 are close enough.
 
Check the Post hoc analysis. 
It indicates in both dep var(s) category 4 is quite different from 5 & 6 (coz Sig values are 4&5 as well as 4&6 are < 0.05)
 
The last tables of Homogenous subsets give same info in different form, so we can skip those tables.
 
Conclusion:
perfo has significant bearing on both dep var web & material.
The perfo category 4 was found to be significantly different from rest.
 
===============================================
 
Obj 2: To check whether web & material together depends on performance while taking Sales as covariate.
 
Dep var: web & material
Indep var: perfo
Covariate: sales
Technique used: One way MANCOVA
 
Go to Analyze->General Linear Model->Multivariate
Put web & metrial in dep var
perfo in Fixed factors
sales in covariate
 
Output Analysis:
Table-2 indicates perfor is related with dep var(s) but sales is not.
Table-3 gives wonderful results. In the presence of sales perfo loose its relationship with webhits (Sig value=0.135) coz sales is nor related with web hits (Sig value as 0.038).
 
Earlier we assumed that perfo is affecting both but now it comes that sales is actually affecting web hits while perfo is affecting CD promo material.
 
================================================
This concludes our discussion on One way MANCOVA.
 
 
Happy learning
Neeraj

[Radha garg]

Respected Sir,

Thanks for completing the story of all ANOCOVA, MNOVA, MANOCOVA finally.

I know its takes a lot of time to teach us through posts. But u done it very well as usual.

So, thanks for ur precious time once again.

 

Cheers

Radha

[Neeraj Kaushik]

Picture abhi baaki hai mere dost .........

 

Repeat measure ANOVA ki kahani bhi to sunani hai :-)

 

Aur abhi Log linear analysis (3 types) bhi to hai pipeline mein.

Stay tuned for more :-)

--
 
 

[Radha garg]

Respected Sir,

Thanks for reminding me such a interesting concepts and hope it have more twist than the previous one. And generalized linear model has also been left.

So, i am eagerly waiting for all.

Regards

Radha