forum.smartpls.com

Hi.
My model has 8 latent variables and 12 control variables.
The control variables are 8 categorical and 4 continuous.
I transformed the categorical control variable to dummy variables.
In result, 22 dummy variables are generated.
So, the number of control variables is 26.(22dummy variables + 4 continuous variables)
In sum, my full model has 8 latent variables and 26 control variables.
When I did Consistent PLS algorithm, "Out of Memory Error: GC overhead limit exceeded" happened. Even though I change the setting of memory parameters(Xmx7168m, Xms1024m), the error was not solved.
Strangely, When I deleted all 22 dummy variables and left only 4 continuous variables, the PLSc algorithm made results.
How can I solve this problem?

Thanks.

How many observations do you have?
Do you only estimate the algorithm or the bootstrapping of the algorithm?

Did you correctly transform categorical into dummy by leaving the reference category out (i.e., if you have 4 categories you have 3 dummies)?

jmbecker wrote: ↑Fri May 25, 2018 1:18 pm How many observations do you have?
Do you only estimate the algorithm or the bootstrapping of the algorithm?

Did you correctly transform categorical into dummy by leaving the reference category out (i.e., if you have 4 categories you have 3 dummies)?

I have 307 samples.
The total number of indicators is 37. (I have 8 latent variables)

I used both Consistent PLS algorithm and Consistent bootstrapping.
In case of algorithm, when I use Consistent PLS algorithm, GC Overhead limit error occurred.
On the other hand, when I used PLS algorithm, the error did not occur and I can get estimate.

In terms of dummy variables, surely, I leaved the reference category out.
When I use Consistent bootstrapping, 0% was done and the same error occurred like PLSc algorithm.
When I deleted some of dummy variables and do regular bootstrapping, I can get results.
But, the result like path coefficient show "n/a" in Sample Mean and Standard Deviation.
I don't know why this kind of problem happened.
Please give me some advice.

Generally, I would say that there shouldn't be a problem, but it could be something specific with you data and model.

Could you please send you data and model to our support: http://support.smartpls.com/
They will investigate the problem.

jmbecker wrote: ↑Sun May 27, 2018 9:31 am Generally, I would say that there shouldn't be a problem, but it could be something specific with you data and model.

Could you please send you data and model to our support: http://support.smartpls.com/
They will investigate the problem.

Ok. I am very thankful for your quick reply.

Best wishes.
KIHO KWON