Errors in Bootstrapping

Frequently Asked Questions about the SmartPLS software. Please refer to this section first, if you have problems with the software.
Post Reply
swoolford
PLS Junior User
Posts: 2
Joined: Fri Apr 13, 2012 4:19 pm
Real name and title:

Errors in Bootstrapping

Post by swoolford » Tue Jul 09, 2013 2:51 pm

I have run the bootstrapping algorithm on a PLS model with both reflexive and formative constructs. The model satisfies the 'rule of 10' (there are 140 observations) but I get the error message:

"Error occured during outside estimation-Maybe there are two few observations"

Has any one experienced this issue?

Tiwelle
PLS Junior User
Posts: 2
Joined: Thu Jul 05, 2012 11:24 am
Real name and title:

Post by Tiwelle » Wed Jul 17, 2013 1:56 pm

Hello,

I have the same problem (sample 90). Did you find a solution ?

Thank you.

swoolford
PLS Junior User
Posts: 2
Joined: Fri Apr 13, 2012 4:19 pm
Real name and title:

Errors in bootstrapping

Post by swoolford » Thu Jul 18, 2013 7:29 pm

I may have discovered the cause but I don't know whether there could be a system or algorithmic explanation. In my case I had a construct in the model that only had one indicator. That indicator only had three values and one value occured roughly 80% of the time. My guess is that one of the bootstrap samples may have only had a single value for this variable and so the OLS estimation failed

User avatar
Diogenes
PLS Super-Expert
Posts: 905
Joined: Sat Oct 15, 2005 5:13 pm
Real name and title:
Location: São Paulo - BRAZIL
Contact:

Post by Diogenes » Fri Jul 19, 2013 11:42 am

Hi,
the Sam's explanation is correct,
if we used the "search" function (green corner), there are other examples, usually with dummy variables (almost constant).

Best regards,
Bido

Tiwelle
PLS Junior User
Posts: 2
Joined: Thu Jul 05, 2012 11:24 am
Real name and title:

Re: Errors in bootstrapping

Post by Tiwelle » Sun Jul 21, 2013 1:00 pm

swoolford wrote:I may have discovered the cause but I don't know whether there could be a system or algorithmic explanation. In my case I had a construct in the model that only had one indicator. That indicator only had three values and one value occured roughly 80% of the time. My guess is that one of the bootstrap samples may have only had a single value for this variable and so the OLS estimation failed
Hi,

Right ! it's the same problem for me.

Post Reply