Hallo,
I am a PLS beginner and I have a question about my model. There is one high path coefficients (0,677) which did not turn out to be significant during Bootstrapping, although there are others with a lower value which are high significant. does that make sense? Or does this finding indicate that I did a mistake?
Thanks,
Claudia
high path coefficients but not significant
-
quant
- PLS User
- Posts: 12
- Joined: Tue Feb 23, 2010 10:06 am
- Real name and title:
- Location: Berlin, Germany
- Contact:
Hi Claudia,
this is not a PLS-specific issue. It is due to the parameter estimate's (which is a random variable!) standard error. The higher the standard error, the smaller the t-value given the same estimated parameter value. Have a look at the "formula" of a t-statistic. If the confidence interval (say the .95 interval) of the parameter contains zero, it is regarded as insignificant, regardless of its value.
Best regards,
Christian
this is not a PLS-specific issue. It is due to the parameter estimate's (which is a random variable!) standard error. The higher the standard error, the smaller the t-value given the same estimated parameter value. Have a look at the "formula" of a t-statistic. If the confidence interval (say the .95 interval) of the parameter contains zero, it is regarded as insignificant, regardless of its value.
Best regards,
Christian
-
stephan.kramer
- PLS Expert User
- Posts: 37
- Joined: Tue Mar 25, 2008 9:16 am
- Real name and title:
Christian is right with his conclusions, although with a path coefficient of 0,677 the standard errors must be very high to prevent this result to become significant. Did you adjust the settings for sample size? You might also consider using a different sign change option: No sign changes is the most conservative method, while individual changes normally reduces standard errors quite significantly.
Best regards, Stephan
Best regards, Stephan