Internal Consistency Factor Loading (item reliabilities) ?

Research topics can be discussed in this area.
Post Reply
HussainWaasly
PLS Senior User
Posts: 21
Joined: Thu Mar 21, 2013 7:37 pm
Real name and title:

Internal Consistency Factor Loading (item reliabilities) ?

Post by HussainWaasly » Wed Nov 06, 2013 8:40 am

Hi All.
I have ongoing research and it is about to finish. I am doing data analysis now and i want to get Internal Consistency Factor Loading Analysis to check item reliabilities.
i have collected data from (Report->PLS->calculation result->outer loading) to get Internal Consistency Factor for each element (composite reliability value).
and from (Report->PLS->calculation result->Outer weight) to get the weight for each elements.

Kindly, confirm me if i am on the right track or no!
Regards,

User avatar
Diogenes
PLS Super-Expert
Posts: 905
Joined: Sat Oct 15, 2005 5:13 pm
Real name and title:
Location: São Paulo - BRAZIL
Contact:

Post by Diogenes » Fri Nov 08, 2013 6:47 pm

Hi Hussain,

The short answer = yes.

The long answer =

If you are using reflective indicators:

1) Convergent validity:
Outer loading = PLS / Calculation results / Outer loadings
AVE = PLS / Quality criteria / Overview

2) Discriminant validity:
Item level --> Crossloading = PLS / Quality criteria / crossloading
Construct level --> (square root of AVE > LV correlations) -->
- compute square root by hand using the AVE from Overview
- put these values in the diagonal of LV correlations matrix

3) Reliability:
Composite reliability and alpha’s Cronbach --> PLS / Quality criteria / Overview

Best regards,
Bido

HussainWaasly
PLS Senior User
Posts: 21
Joined: Thu Mar 21, 2013 7:37 pm
Real name and title:

Post by HussainWaasly » Tue Nov 12, 2013 7:12 am

Thanks a lot for the usual support!

to test "Research Hypotheses" i need to check path coefficient t-values (bootstrapping/path coefficient (Mean,STDEV,t-values)) if it is more than 1.96 then i assume that the path coefficient significantly different from 0 at a significance level of 5 percent (alpha = 0.05; two-sided test) .
and Research Hypotheses supported!
Don't I?

what about R2 and how can i get R2 for the full model?

I am trying to do the assessment of the structural model. I know that R2 should be more than 2 but it is for what?

there are two more steps :
Report / PLS / Calculation results / Path coefficients
- The same results that were showed in the model
but what does it mean?
Report / PLS / Quality criteria / Total effects
- you could assess the direct, indirect, and total effects.
also but what does it mean?

Regards,

User avatar
Diogenes
PLS Super-Expert
Posts: 905
Joined: Sat Oct 15, 2005 5:13 pm
Real name and title:
Location: São Paulo - BRAZIL
Contact:

Post by Diogenes » Tue Nov 12, 2013 9:55 am

Hi Hussain,

1) Yes, t > 1.96, path is significant, hypothesis supported.

2) Each endogenous LV has a R²
Endogenous are LV that received structural arrows.

You could think in regression terms.
Y = b1*x1 + b2*x2 + b3*x3
Here, we will have a R² for Y regression.


3) Yes path coefficients from report = model (image)

4) Direct, indirect and total effects should be assessed if your model has indirect effects.
Example:

a) X1 –>0.4--> X2 -->0.3  X3

b) X1 --> 0.2  X3

From (a): X1 has a indirect effect on X3 = 0.4 * 0.3 = 0.12
From (b): X1 has a direct effect on X3 = 0.2
Indirect + direct = total effect = 0.32

SmartPLS shows direct effect in the model and report as path coefficients, and total effect in the report, if you want to compute indirect effect, just subtract one of other.

Best regards,

Bido

HussainWaasly
PLS Senior User
Posts: 21
Joined: Thu Mar 21, 2013 7:37 pm
Real name and title:

Post by HussainWaasly » Tue Nov 12, 2013 11:23 am

Thanks again Prof.
small questions to make sure every thing in place.

if R2 > 2 or R2<2 , what does it mean?
and to get R2 for full model, i have to calculate all R2 for all LV to get R2 for the full model. please advise ?


Regards,

User avatar
Hengkov
PLS Super-Expert
Posts: 1619
Joined: Sun Apr 24, 2011 10:13 am
Real name and title: Hengky Latan
Location: AMQ, Indonesia
Contact:

Post by Hengkov » Sun Nov 24, 2013 12:38 am

Hi,

R^2 for variance explain each endogenous construct. We recommended to report Adjusted R^2. According to some literature, R^2 > 0.25 is good. If you interest to know R^2 model, you must compute terms ARS (Average R^2): Sum all result R^2 in model/number of endogenous variables.

Best Regards,

Post Reply