Hi,
I know that if one would like to test if one path coefficient (e.g. b1) is bigger than another (b2), usually following formula can be used:
t=b1-b2/(varb1+varb2-2covb1b2)^0,5
However, I also need to determine the degree of freedom (DF) for the t-test. Now my question is how can I obtain degree of freedom?
In usual t-test, the degree of freedom is the sum of the N in both groups minus 2. Is it applied to this context?
For example, if I used bootstrapping with 500 resamples, DF would be 500 (NUMBER of bootstrap sample 1) + 500 (NUMBER of bootstrap sample 2) -2 = 998.
Is it right?
Taeshik
comparison of path coefficients
-
Taeshik Gong
- PLS Junior User
- Posts: 1
- Joined: Thu Feb 15, 2007 1:12 am
- Real name and title:
-
Diogenes
- PLS Super-Expert
- Posts: 899
- Joined: Sat Oct 15, 2005 5:13 pm
- Real name and title:
- Location: São Paulo - BRAZIL
- Contact:
Hi,
1) In each sample the value of b1 is computed based in n cases, even been computed 500 times b1 is computed using n cases.
2) When you run the bootstrpping procedure, you have 500 values of b1, and its mean and standard deviation.
3) Maybe a t-test for means (Ho: b1 - b2 = 0) using this results (500 values of b1 and 500 of b2) could solve the problem.
(Or non-parametric version: Mann-Whitney or Kruskal-Wallis).
I understand that the formula that you presented is used with one sample and n cases.
Other explanation in: http://disc-nt.cba.uh.edu/chin/plsfaq/plsfaq.htm
Best regards.
Bido
1) In each sample the value of b1 is computed based in n cases, even been computed 500 times b1 is computed using n cases.
2) When you run the bootstrpping procedure, you have 500 values of b1, and its mean and standard deviation.
3) Maybe a t-test for means (Ho: b1 - b2 = 0) using this results (500 values of b1 and 500 of b2) could solve the problem.
(Or non-parametric version: Mann-Whitney or Kruskal-Wallis).
I understand that the formula that you presented is used with one sample and n cases.
Other explanation in: http://disc-nt.cba.uh.edu/chin/plsfaq/plsfaq.htm
Best regards.
Bido