forum.smartpls.com

ALL indicators in PLS are formative, whether the outer, "measurement" model is specified as being inner or outer directed. All that varies is the estimation method by which the weights are determined. To avoid the problems you describe (due to multicollinearity) you should specify your out...

That's actually one of the key points about PLS--it does not make any distributional requirements on the data.

The classic example of a construct with formative indicators is Socio-Economic-Status, which is very analogous to your Economic Success construct. So yes, creating a latent variable from multiple "objective" measures is done. As an aside, two points: 1) Even "objective" measures ...

Any chance of making the missing data value something other than -999? It has never happened to me, but I always wonder about someday using some dataset where -999 is a valid potential data value! Could it become some non-numeric symbol? A period, ala SAS is good, or any others (e.g. !,@,#,$,%,^,&am...

In your example, where the distributions in the dataset are Brand A 75% and Brand B 25% but the actual market shares are 50% and 50% you would weight the observations as follows: Each Brand A respondent would have a weight of .667 applied to it (so now each Brand A respondent is counted as .667 of a...

Let's imagine that we have data from owners of products in a particular category, Brand A, B, and C. In our data, the distribution of ownership is as follows: Brand A 15% Brand B 50% Brand C 35% However, the brands have the following market shares: Brand A 25% Brand B 35% Brand C 40% Especially in a...

Great question! If it isn't implemented in this release that would be a wonderful feature for the next. But I'm not sure about boostrapping with a variable weight. The variable weight, for example, might be to make the data more representative (ie., your data is skewed sucht that there are too many ...

And, at least on my system, after the final click, with either CNTRL or SHIFT depressed, without releasing the click, move the indicators to the latent variable. If you, for example, click on one indicator, then CNTRL click on several others, then let go of the mouse button and try to move the group...

The problem, I think, is that such PLS regression does not really address the multicollinearity at the level of your latent variables and shares many (but not all) of the problems that plague Principle Component Regression. On the positive side you do not have the problem of the arbitrary number of ...

Yes! As they say in Chicago, I may have to vote early and vote often on this one! Going even a step beyond total STRUCTURAL effects, there have been times when I have been interested in total effects from indicators to some ultimate dependent variable of interest (of course this is simply taking the...

My apologies, I wasn't following the model you were describing. I assume you're measuring the second order latent variable with all of the indicators of the first order latent variables? If so, then what I was suggesting should still work, I think. specify the measurement model for each latent varia...

Also, consider what the multicollinearity of the indicators is telling you: perhaps the construct should not be measured via formative indicators. It would seem more suited to a reflective, true score model (ala LISREL). In this case, the only rationale I can see for using PLS (becuse, again, PLS la...

Can you explain more what you mean by PLS regression in the structural model?

When I create second order latent variables--in fact, for all latent variables, first order or second order--I typically set the outer model to be outer directed. Latent variables in PLS are always formative composities, regardless of how the outer model is specified (McDonald, 1996). As long as the...

A great future feature would be the ability to generate Monte Carlo data: specify an inner model and an outer model and have the program generate data sets based on a specified distribution around the true values.