Two datasets with same variables but different results

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
Post Reply
Marius.Huber
PLS Junior User
Posts: 1
Joined: Sat Sep 05, 2020 9:55 am
Real name and title: Marius Huber, PhD

Two datasets with same variables but different results

Post by Marius.Huber »

Dear all,

recently I ran into a problem which I am not able to solve myself.

I have two CSV files that I use as datasets. Both have the same data from the same data collection with n = 129.
But one of them consists of only three variables each measured with two items and the other dataset includes a few more variables, including the same variables that I mentioned first.
Currently I am trying to measure a model that is based on these three variables.
The model has one IV, one mediator and one DV. Each variable measures the same construct but in different timestamps.

My problem is that these two files/ datasets produce different results.
The dataset that holds more variables shows issues in terms of reliability (e.g. rho_A = 3.450, CR = 1,683), validity, path coefficients (negative instead of positive paths) and p-values whereas the smaller one shows no such problems.
However, in SmartPLS and Jamovi I had a look at the descriptive statistics for those three variables and they are the exact same.
In terms of variables, models and settings everything stays the exact same (PLS algorithm and bootstrapping) but Smart PLS (V 3.3.9.) produces different results.

I assume that if the descriptive statistics are the same, PLS should produce the same results for that model when the variables, model and settings stay the same or am I wrong?

The model will be later extended with the other variables collected but for now both datasets should produce the same results.

Greatly looking forward receiving help on this topic.
If any more info is needed, please let me know.

Best,
Marius
Post Reply