forum.smartpls.com

I want to force (constrain) all the indicators for a formative factor to have positive (or, perhaps better said, non-negative) weights.

Can I do this with SmartPLS ?

I almost always filter out cases with missing values using the routine in SmartPLS. Can the resulting sample size be displayed? So far as I can tell, this is not included in the report either. (But, as always, I might be wrong.)

I may have this wrong, but can the next version include the option to simultaneously display the weights/coefficients/etc _and_ the t statistics (from bootstrapping) ?

I managed to have a non-recursive component in my model by including the same factor and its indicators _twice_ ... i.e. as two separate constructs. That way I could have an arrow going in one direction from the first factor to a third factor and an arrow going in the other direction from the third ...

I have a file with lots of missing values, and have used the casewise deletion option. Apologies for the (dumb) question, but where in the output do I find the number of cases that were actually used in the analysis ?

Just make sure that you have not formatted your ID variable to have a comma separator when it gets into the 1000's ... otherwise the package will think you have one more column of data than you actually have.

Not infrequently, when running a model off a big database (.csv file) the calibration process seems to halt and nothing happens, or else it simply doesn't even get to iteration zero. This is after I have requested lots of memory via the options. Invariably, however, I can fix this simply by closing ...

I was having a similar problem as yours, until I stopped keeping my csv files in the same directory as the project directory.

[In my case, the import not only showed no data, it also deleted the csv file altogethre !!]


Hope that helps.

I think this may be a major bug - if I save the data (.csv or .txt) into the same directory as the project, then when importing the data it actually shows no data in the validation window and it also deletes the datafile from the directory. By experimenting, it seems that by having the datafile NOT ...

This option would also solve the problem of not currently being able to analysis weighted data sets, since the variances/covariances can be computed on a weighted basis. I have used this trick with AMOS to enable weighted models to be run (although what it does to the interpretation of significance ...

If you delete the variables/paths you don't want in the model, then run it, you can inspect the results and then get those variables and paths back again with the <Undo> command, i.e. <ctrl> <z>. This way, you don't have to redraw part of the model.

Hello, I have a question about running the same model many times with different datasets. I have a huge data set with one classification variable and it is lots of work to manually run the software for every value of the class variable. Is there any way to automatize the Smart PLS software, so that...

Well, it did turn out to be a memory problem after all. This time, on starting up I went straight in to SmartPLS without loading any other programs (SPSS, Word, Excel etc) and this seemed to leave enough memory available for the non-running model to run. Plus ... I have now upgraded to 4mB RAM inste...

alan_scott wrote: t looks like it might be a memory problem so I will try increasing the allocation:

No, that didn't work :-(

When I delete one of the factors altogether, it runs OK. But that same factor is included in the other model without any problems. So I still don't know what is going on.

Scott

I have two databases (.CSV format) and models which are almost identical except for two variables. Both CSV files validate OK. One model runs OK and quickly. the other one does not show a progress bar and produces no results. The problem is, because there is no output when the model doesn't run (unl...