Pseudo-replication-sample size

This forum is closed, and read-only.
Locked
alizol
PLS User
Posts: 11
Joined: Wed Nov 02, 2016 5:54 pm
Real name and title: Ali Zolfaghari

Pseudo-replication-sample size

Post by alizol »

Dear professors,

I have a question about Pseudo-replication.

Imagine that I want to test the fitness of a model for measuring performance of new product development projects. I consider the total performance of new product development (TTNPD) as a second order construct which have -for example- two dimensions which are first order constructs. Let's name them as strategic performance (SNPD) and technical performance (TNPD) which have their own indicators as showed in the attached model (word file attached).
As an exemplification please imagine that I have two development projects which have different conditions -and there is a consensus between experts that they are independent from each other- and I have ten experts to judge about different indicators of first order constructs (two dimensions).

1- Could I multiply ten to two (the no of experts*the no of projects) and say I have twenty samples which means that the sample size for bootstraping is twenty? . The model is based on theory and not made by myself. I don't want to say something about similarity of PLS parameters estimates' so I forgot about multigroup analysis. I just want to test the model fitness with SmartPLS2 and I don't want to do any inferential test(like t-test or annova...).

2-Some say that it is Pseudo-replication and not allowed in PLS.But they don't say why it is so considering the independence of projects and experts from each other.If this is Pseudo-replication could I take an avearge from 2 projects for every 2 experts and build 10 SPSS records to import in SmartPLS2.

Sincerely Yours
Thanks in Advance
alizol
PLS User
Posts: 11
Joined: Wed Nov 02, 2016 5:54 pm
Real name and title: Ali Zolfaghari

Re: Pseudo-replication-sample size

Post by alizol »

(Writing correction)
Dear professors,

I have a question about Pseudo-replication.

Imagine that I want to test the fitness of a model for measuring performance of new product development projects. I consider the total performance of new product development (TTNPD) as a second order construct which have -for example- two dimensions which are first order constructs. Let's name them as strategic performance (SNPD) and technical performance (TNPD) which have their own indicators as showed in the attached model (word file attached).
As an exemplification please imagine that I have two development projects which have different conditions -and there is a consensus between experts that they are independent from each other- and I have ten experts to judge about different indicators of first order constructs (two dimensions).

1- Could I multiply ten to two (the no of experts*the no of projects) and say I have twenty samples which means that the sample size for bootstraping is twenty? . The model is based on theory and not made by myself. I don't want to say something about similarity of PLS parameters estimates' so I forgot about multigroup analysis. I just want to test the model fitness with SmartPLS2 and I don't want to do any inferential test(like t-test or annova...).

2-Some say that it is Pseudo-replication and not allowed in PLS.But they don't say why it is so considering the independence of projects and experts from each other.If this is Pseudo-replication could I take an avearge from 2 projects for every expert and build 10 SPSS records to import in SmartPLS2.

Sincerely Yours
Thanks in Advance
Locked