Slow Bootstrapping

edemir · Post by **edemir** » Fri Jul 06, 2007 9:06 am

Hi,

the bootstrapping procedure on my PC is extremely slow..last time, I let it run over night but without success. The process was not finished since in the morning. I do not have any idea what should I do.

My sample size accounts 2139 companies with each 40 questions, and not every question is answered (I do not know how many questions are not answered due to large sample size. Should I know it? and if how?).

I set missing values and case wise replacement. As I have read, the cases number should be as high as the original sample size, so I set it to 2139. Sample value is 3000. Is that correct?

In addition, I increased in the preferences the required memory to 1000000 bytes (1 MB). My notebook has 512MB. But that doesn't help to increase the speed of bootstrapping. What should I do?

It would be great if you could help me. Thank you. Emine

jmbecker · Post by **jmbecker** » Fri Jul 06, 2007 10:11 am

You may use the search function in this Forum for the term "memory". There are a lot of topics concerning this fact.

My first suggestion is that you should try to increase the memory to at least 256MB or maybe even more.

edemir · Post by **edemir** » Sat Jul 07, 2007 6:58 am

Hi, thank you for your response. I will check this. Emine

edemir · Post by **edemir** » Sat Jul 07, 2007 8:22 pm

Hi,

no way...I have increased both the memory in system settings and from the Java Virtual Machine. Just little faster the bootstrapping runs and I still need the whole night. Is that normal, if I have only 512MB memory space and a sample size of about 2000?

Sorry for the several questions, but I have no much time for my thesis and it is important that the bootstrapping is more faster. Many thanks. Emine

edemir · Post by **edemir** » Mon Jul 09, 2007 7:28 pm

Hi,

now I get the opportunity at my chair to increase the memory to 1.5 GB. But the bootstrapping procedure also is cancelled at 50 samples. Has anyone experience with so high sample sizes (cases 1332, sample 2000)?

I will be very thankful for any information.

Emine

jmbecker · Post by **jmbecker** » Tue Jul 10, 2007 10:29 am

Hi,
there must be something strange with your model.
I just tried an example with about 3000 cases (9 LVs with 5 to 7 indicators); 2000 samples.
All samples went trough in less than an hour!

(btw. Java Memory assigned: 512MB)

edemir · Post by **edemir** » Tue Jul 10, 2007 12:49 pm

hi, thank you.

Now I just tested a smaller model with only three LV and a few indicator. I get the same error pop up at 50 samples. That is strange.

My original model include 10 latent exogenous variable and one latent endogenous variable which is modeled to be formative with 26 indicator. The latent exogenous variable are operatinalized differently, 7 formative and 3 reflective.

Where may be the problem?

Cheers,
Emine

edemir · Post by **edemir** » Thu Jul 26, 2007 11:56 am

Hi,
i just wanted to inform that I have run a new installation and now everything is well.

Thank you anyway

zhangx · Post by **zhangx** » Sat Sep 29, 2007 3:51 am

Hi, Emine ,

Can you describe how you model an exogenous construct with 7 formative and 3 reflective constructs? I tried in SmartPLS but I can only create either all reflective or all formative indicators for the same exogenous construct.

jmbecker · Post by **jmbecker** » Tue Oct 02, 2007 9:38 am

She had overall 10 exogenous variables, which were either formative (7 of them) or reflective (3 of them).

forum.smartpls.com

Slow Bootstrapping

Slow Bootstrapping

mixed measures for the same exogenous contruct