Interpreting the relative impact of formative Indicators

[saab]

Hi all,

I have the following question: How can the relative impact of formative Indicators of a LV be interpreted?
Let's say I have a LV with three formative MV's, each with a positive and significant weight.
The weights are: 0.25 (MV 1), 0.30 (MV 2), and 0.35 (MV 3), just for example. I might intepret this result as "MV 3 has the strongest influence on the LV".
However, is the order of the MV's really statistically significant? How can I estimate, that the difference between the calculated weights is significant?

Thanks for your ideas in advance!

-Samy

[Diogenes]

Hi Samy,

I thought a way:
1) Run bootstrat
2) Take the standard deviation to compute a I.C. (usually +/- 3 S.D.)
3) If the another weight is in this I.C., probably the difference isn´t significant

(in your example: 0,25-0,30-0,35 will be statiscally different if the S.D. was smaller than 0,05/3)

Just a idea, but could help you.

Best regards.

[Diogenes]

Hi,

I´ve found a clue in
YUNG, Yiu-Fai; CHAN, Wai. Statistical analysis using bootstrapping: concepts and implementation. IN Hoyle, Rick H. Statistical strategies for small sample research. Sage Publications: 1999, p. 81-105.

1) Bootstrapping with 1000 or 2000 or 3000 replications
2) Use +/- 3 SE or 1.96 SE (95%), in the example they was comparing alpha de Cronbach.
3) or take the 2.5th and 97.5th percentiles as the critical points

Best regards