|
| Algorithm Parameter Settings Wizard |
We enhanced usability of parameter settings. The settings button in the task bar is removed. Instead, a parameter settings dialog pops up for every SmartPLS computation.
 |
|
| Moderating Effects |
We implemented a tool to support the product indicator approach for moderating effects proposed by Chin, Marcolin and Newsted . We recommend to read their paper before you start modeling interaction effects with SmartPLS.
Modeling the interaction term for this kind of analysis is fairly easy. Please follow the instructs below or take a look at this video.
Just hit the context menu on the dependent construct and choose "Create Moderating Effect".
Having reflective measurement model in both, predictor and moderator variable, the indicator variables of the predictor and moderator constructs are used to generate new product indicators. These product terms measure the additionally created interaction term in a reflective measurement model (see figure below). According to the paper by Chin, Marcolin and Newsted it is recommended to mean-center or standardize the product indicators. Graphically moderating effects are presented as follows: |
If you change the measurement model of the predictor and/or the moderator variable AFTER you created a moderating effect, it is critical to alter the product terms and the measurement model of the interaction term. This can be a nasty and time consuming job. Unlike other software applications, SmartPLS automatically updates the product indicators according to changes in the measurement model of the predictor and/or moderator variable. JUST TRY IT!
Important note: The paper by Chin, Marcolin and Newsted states, that the product indicator approach should only be applied in case of reflective modeled constructs. Remember this, when you model your constructs. SmartPLS does not restrict the modeling in this case to provide the ability to explore the behavior of this technique. |
|
| New: Blindfolding |
We implemented the blindfolding procedure that can be started from the calculate menu.
The procedure must be parameterized via the following property page. Apparently there is only a single parameter, which allows for changing the omission distance.
Note: The number of cases in your data must not be a multiple integer number of the omission distance (otherwise the blindfolding procedure cannot be carried out).
The report for the blindfolding results contains the cross validated communality and the cross validated redundancy for constructs as well as indicators.
|
|
| PLS-Report: Cross loadings, Communality, Redundancy, Cronbach's Alpha |
We added requested quality criteria to the PLS results report:
- Cross loading
- Communality
- Redundancy
- Cronbach's Alpha
Cronbach's Alpha can also directly be displayed as a result in the graphical model (this is a display option in the preferences). |
|
| Workspace Selector |
SmartPLS no longer depends on a fixed workspace directory for all your SmartPLS projects. Instead SmartPLS asks you to select a workspace directory during the startup routine. This enables you to manage different workspace on your computer.
Note: It is possible, to zip and transfer workspace directories with all your projects from one computer to another!
The following dialog appears at the first SmartPLS startup. It allows you to specify the workspace directory that you like to use for the current SmartPLS session. The dropdown menu contains your recent workspaces. |
If you do not want to manage different workspaces, you can remember the current location as your personal default by checking the "use this as the default and do not ask again" option. (To re-enable the startup prompt: use the preference settings, which are accessible via the Window -> Preferences menu - see the following screenshot).
 |
|
| New: Outline Window |
We implemented an additional window with an outline perspective that can be used to navigate large models. The view shows a tree structure of the current path model. Selecting variables in this view also selects the variables in the path model, deleting variables in this view also deletes variables in the path model. Calculation results are also displayed in the outline view. Please also check the context menu of this view for further options.
|
|
| Indicators Window |
We changed the presentation of the manifest variables in the indicators window. Manifest variables that are already used in the current path model will be highlighted in yellow/black color. Unused manifest variables will be appear in black/white color.
For a better overlook during modeling you can show/hide indicators that are already used in the current model using the  button.
|
|
| Improved Report Editor |
| We improved the presentation of calculation results in the integrated report editor. Captions of variables are no longer truncated. Additionally variables are sorted in alphabetical order in all result matrices.
Note: All results can be easily transferred via copy and paste to other software applications (e.g. MS Excel). |
 |
|
| Bootstrapping |
It is now possible to display the t-values as a bootstrapping result directly in the path model.
In addition, we corrected some implementation problems. In detail, the following bugs have been fixed:
- problems with the random number generator induced double selection of the same cases on very fast computers,
- the chosen sign change options were not applied during calculation,
- t-values were not correctly calculated .
Note: These changes do not seriously alter the bootstrapping results compared to the previous SmartPLS version. The only exception are mixed sign measurement models path models that were bootstrapped with the construct level sign option. |
|
| Improved Data Editor |
We improved the user interface for the data editor that can be used to configure the correct data format for imported data files.
Note: Additionally, missing values should be configured via this editor, too ! |
 |
|
| Model Editor |
Preferences
We implemented a preference page, which can be used to customize functionalities of path modeling editor window. Currently, there are the following settings:
Shortcuts
Furthermore, new shortcuts for the three editor modes allow for faster modeling:
- selection mode = STRG-1,
- insert mode = STRG-2,
- connection mode = STRG-3.
Alignment behavior
We corrected the behavior of aligned indicators. When adding an indicator to a latent variable that already has aligned indicators using one of the four alignment options (top, bottom, left, right), the additional indicator is aligned in the same direction.
Furthermore the indicators will now be aligned in alphabetic order by default (see also the according preference setting).
Renaming variables
We replaced the in place editing component by a dialog due to some problems with larger path models.
Hidden indicators behavior
Some indicators displaying problems (indicators being blinded out) are corrected. In addition, hidden indicators are now signalized by a bold caption of the construct's name.
Improved model presentation
As you can see in the following screenshot, we improved the model presentation. The caption of latent variables is now placed outside the circle and will change its position according to the chosen indicator alignment.
Presentation of weights and loadings after applying the PLS algorithm
According to the construct type (formative, reflective) either weights or loadings will be displayed.
|
|
New: Report Menu |
We provide a new report menu, which allows you to select the output format for the reports. The menu is also available from the toolbar. There is a new Microsoft Excel report type available, which is still experimental.
|
|
| Bugfixes |
- Titles of all editors show the correct file names.
- Because of heavy-handed shortcuts, some letters could not be written in upper case in the model editor (fixed now).
- Some headers in the bootstrapping result matrices were permuted (fixed now).
- Moving selected constructs and indicators caused errors in some cases (fixed now).
- Folding indicators caused errors under certain circumstances (fixed now).
- The project creation wizard caused errors, when a empty project name was entered (fixed now).
- Removed dead entries in the help menu (fixed now).
- Undo function crashed under certain circumstances (fixed now).
- Starting a new calculation did not clear result values displayed in the graphical model under
certain circumstances (fixed now).
- Missing value processing is activated by default now .
- Missing value processing was unnecessary applied to files that had no missing value configured (fixed now).
- Copying an active data into the source project of the file caused errors (fixed now).
- Customized report label settings were not applied (fixed now).
- Old result reports were cached, which lead to outdated results in the report view under certain circumstances (fixed now).
- Corrected calculation of bootstrapping t-values.
- Bootstrapping sign changes settings were not applied (fixed now).
- Problems with the random number generator induced double selection of the same cases
on very fast computers (fixed now).
- Problems with missing GDI library on Windows (fixed now).
- Experimental t-values from PLS report are removed.
- Hitting the [Cancel] button during the calculation of bootstrapping and FIMIX-PLS caused errors (fixed now).
- The calculation of large models caused out of memory errors (fixed now).
|
