Great differences of bootstrapping result in ver 2.0 and 3.0

[choiswf]

Hello,

I've found that the bootstrapping results of moderating effect in SmartPLS 2.0 are quite different from those generated by SmartPLS 3.0.

The fact is I've first created a model with moderating effect in SmartPLS 2.0 and then exported and imported it to SmartPLS 3.0, so the model should exactly be the same.

I've captured images of bootstrapping results from both versions (see below). Can anyone help explain the great differences found?

Below are my settings of bootstrapping in SmartPLS 2.0 and SmartPLS 3.0.

Bootstrapping settings in SmartPLS 2.0
- Interaction effect term generation: use original indicator values for multiplication
- 1,000 subsamples
- No Sign Changes

Bootstrapping settings in SmartPLS 3.0
- Product Term Generation: Unstandardized
- Calculation method: Two Stage
- Weighing Mode: Automatic
- 1,000 subsamples
- No Sign Changes
- Amount of Results: Basic Bootstrapping
- Bias-Corrected and Accelerated (BCa) Bootstrap
- One Tailed test type
- Significance level = 0.05

Many thanks,
Stella

[jmbecker]

The simple answer is: Don't use SmartPLS 2 for moderating effects. They are not correct.

Since the release of SmartPLS 3, if you download SmartPLS 2 you will get the following message:
Support for SmartPLS 2.0.M3 ended
SmartPLS 2.0.M3 was released in 2005. Since then there has been no maintenance or bugfix. Although it is very popular, SmartPLS 2.0.M3 contains some known issues that have never been fixed. You need to know these issues and some expert knowledge applying the necessary work arounds. SmartPLS 2.0.3 requires an out-dated Java Runtime Environment in Version 6, which has serious security problems and is no longer supported by Oracle. You will not be able to run SmartPLS 2.0.3 on Mac OSX. Although we continue providing free downloads and licenses for SmartPLS 2, we do not offer any support.

There are several advancements in SmartPLS 3 that correct issues around moderating effects in SmartPLS 2. For example, the interaction term is standardized in SmartPLS 2, but should not be. Also having missing values in moderation models provides incorrect results in SmartPLS2.