Low R2 and AVE score

Research topics can be discussed in this area.
Post Reply
nle043
PLS Junior User
Posts: 5
Joined: Fri Jan 31, 2014 8:45 am
Real name and title:

Low R2 and AVE score

Post by nle043 » Thu Feb 06, 2014 4:45 pm

Hi,
I've already run my PLS Algorithm but my AVE score is below than 0.5. Only 2 out of 7 items have score more than 0.5. My CR score is OK but my R2 is low ranging from 0.04-0.29 and there is one blank R2 score. My questions are:
1. Why my AVE and R2 score are low (including blank R2 score) and how to increase it?
2. Did I miss any steps during data preparation in SPSS? I've done:
(a) Missing data analysis.
(b) Check for normality: skewness and kurtosis.
(c) Check for outliers: z-values and Mahanalobis distance.
(d) Descriptive analysis.
3. How do I know which questions to remain and which questions to remove? When and how to do this?

This is the first time I used SmartPLS and I learn to use it through manual, reading and YouTube. I don't have any formal training yet as the next available course is in April.

Thanks in advance.
Lela

User avatar
ghozali
PLS Expert User
Posts: 44
Joined: Sat Oct 15, 2005 1:18 am
Real name and title: Prof. Imam Ghozali, Ph.D
Location: Indonesia

Post by ghozali » Sat Feb 08, 2014 4:18 pm

AVE less than 0.50 is due to your loading factor less than 0.70. Since multiple R squares (R^2) is equal to loading factor^2 (power 2). If your loading factor equal to 0.70 then your R^2 = (0.7^2) =0.49 equal to AVE


Prof. Imam Ghozali
Diponegoro University, Indonesia
Faculty of Economics, Diponegoro University
Jl. Erlangga Tengah 17 Semarang, Indonesia
ghozali_imam@yahoo.com

herotho
PLS Junior User
Posts: 7
Joined: Thu Sep 26, 2013 10:22 pm
Real name and title:

Post by herotho » Thu Feb 13, 2014 5:11 pm

hi,
only look your AVE, then delete one indicator. Then check if your AVE is >0,50, if not delete another indicator. This in reflective constructs.
The literature say that you can conserve the indicator if outer loading >0,40, only do elimination if the AVE is <0,50, or Composite reliability is <0,708; (Hair et al.,2014).

In formative constructs, conserve the items that have Outer loadings over 0,50; only if theory say that its an important indicator. So you can conserve indicators with non significant outer weight. Use the excel to convert t-values to p-values. You obtain t-values from bootstrapping procedure, in SmartPLS. if p-value<0,05 the indicator have significant influence in the construct. the magnitude of the influence is another thing, and the outer weight give you that information.



Why you need increase R2?,
only if you are doing redundancy analysis you need R2 over 64% (correlation between constructs >0,80). redundancy analisis is relate two constructs (one formative and other reflective) that are measuring the same concept.

If it´s not a redundancy analysis, i think you dont need high R2, only need significant correlation between the constructs.

PLS-SEM is a non-parametric analysis, i think you dont need check normality. Only if you are calculating correlation, because you need to know if use Pearson or Spearman correlation analysis.


what is the mahanalobis distance ??

nle043
PLS Junior User
Posts: 5
Joined: Fri Jan 31, 2014 8:45 am
Real name and title:

still a bit unsure

Post by nle043 » Thu Feb 13, 2014 8:05 pm

Hi..
Thanks for your help. I have tried it and I think it works. I am just a little bit unsure about my formative construct as it is still low. <0.5 (AVE) and it has 2 values. For instance, MOTIVATION, I have pull and push values in outer loading score but I also has Motivation values. which one should I delete? the values on pull and push or motivation? and why do I have AVE's score for all three? Thanks in advance.

Regards
leyla
herotho wrote:hi,
only look your AVE, then delete one indicator. Then check if your AVE is >0,50, if not delete another indicator. This in reflective constructs.
The literature say that you can conserve the indicator if outer loading >0,40, only do elimination if the AVE is <0,50, or Composite reliability is <0,708; (Hair et al.,2014).

In formative constructs, conserve the items that have Outer loadings over 0,50; only if theory say that its an important indicator. So you can conserve indicators with non significant outer weight. Use the excel to convert t-values to p-values. You obtain t-values from bootstrapping procedure, in SmartPLS. if p-value<0,05 the indicator have significant influence in the construct. the magnitude of the influence is another thing, and the outer weight give you that information.



Why you need increase R2?,
only if you are doing redundancy analysis you need R2 over 64% (correlation between constructs >0,80). redundancy analisis is relate two constructs (one formative and other reflective) that are measuring the same concept.

If it´s not a redundancy analysis, i think you dont need high R2, only need significant correlation between the constructs.

PLS-SEM is a non-parametric analysis, i think you dont need check normality. Only if you are calculating correlation, because you need to know if use Pearson or Spearman correlation analysis.


what is the mahanalobis distance ??
Lela

herotho
PLS Junior User
Posts: 7
Joined: Thu Sep 26, 2013 10:22 pm
Real name and title:

Post by herotho » Thu Feb 13, 2014 8:54 pm

hi,
formative constructs dont have AVE values. So check if the arrows flows from the indicators to the construct.

[/img]

Post Reply