Hi, there
Is it usual to develop the formative or reflective indicators with the FactorAnalysis from e.g. 3-5 items of my survey? Or do you just use the indicator as an item directly?
Thanks for your help
development of the indicators
-
Philipp Hartig
- PLS Junior User
- Posts: 2
- Joined: Sat Dec 31, 2005 2:03 pm
- Real name and title:
-
stefanbehrens
- PLS Expert User
- Posts: 54
- Joined: Wed Oct 19, 2005 5:53 pm
- 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,
We have both possibilities:
1) When our research is exploratory (there is little information about what variables could be used like indicators of LV), we could use exploratory factor analysis (common factor if we are thinking variables as reflective indicator or principal components if we are thinking variables as formative indicators) to find what variables will be used in future researches or to model structural relations (relations between LV) with another samples (it’s recommended).
Remember that if it’s possible (or expected) that LV covariate it’s better using oblique rotation (not orthogonal like Varimax), but sometimes the results with different methods of extraction and rotation produces the same factor structure (academic versus practical issue).
Another thing, in exploratory factor analysis all variables have factor load in all factors (some of them is high), but in SEM models the variables have loads just in yours LV (zero in the others LV).
In p. 17 Eberl (2004) presents the C-OAR-SE process (by Anlehnung an Rossiter, 2002) to specify the measurement model:
Construct definition: object, attribute
Object classification: open ended interviews, generation of item parts to represent the object
Attribute classification: open ended interviews, generation of item parts to represent the attribute ==> reflective or formative
Rater identification: individual, experts, group
Scale formation: Combination of attribute and object item parts, pretest
Enumeration: total scale scores derived by indexes and averages
EBERL, Markus. Formative und reflektive Indicatoren im Forschungsprozess: Entscheidungsregeln und die Dominanz des reflektiven Modells. Ludwig-Maximilians-Universität München, 2004.
Available on http://www.imm.bwl.uni-muenchen.de/inde ... &Itemid=48
Another good reference about scale construction using factor analysis is PETT, Marjorie A.; LACKEY, Nancy R.; SULLIVAN, John J. Making sense of factor analysis: the use of factor analysis for instrument development in health care research. Sage Publications, 2004.
2) If we are working in the test of theory context, we already know what variables represent (form or reflect) each construct from previous research. For this reason we could use the variables directly (with help of the theory, previous research and your knowledge about the LV and relations).
I hope it helps...
Best regards
We have both possibilities:
1) When our research is exploratory (there is little information about what variables could be used like indicators of LV), we could use exploratory factor analysis (common factor if we are thinking variables as reflective indicator or principal components if we are thinking variables as formative indicators) to find what variables will be used in future researches or to model structural relations (relations between LV) with another samples (it’s recommended).
Remember that if it’s possible (or expected) that LV covariate it’s better using oblique rotation (not orthogonal like Varimax), but sometimes the results with different methods of extraction and rotation produces the same factor structure (academic versus practical issue).
Another thing, in exploratory factor analysis all variables have factor load in all factors (some of them is high), but in SEM models the variables have loads just in yours LV (zero in the others LV).
In p. 17 Eberl (2004) presents the C-OAR-SE process (by Anlehnung an Rossiter, 2002) to specify the measurement model:
Construct definition: object, attribute
Object classification: open ended interviews, generation of item parts to represent the object
Attribute classification: open ended interviews, generation of item parts to represent the attribute ==> reflective or formative
Rater identification: individual, experts, group
Scale formation: Combination of attribute and object item parts, pretest
Enumeration: total scale scores derived by indexes and averages
EBERL, Markus. Formative und reflektive Indicatoren im Forschungsprozess: Entscheidungsregeln und die Dominanz des reflektiven Modells. Ludwig-Maximilians-Universität München, 2004.
Available on http://www.imm.bwl.uni-muenchen.de/inde ... &Itemid=48
Another good reference about scale construction using factor analysis is PETT, Marjorie A.; LACKEY, Nancy R.; SULLIVAN, John J. Making sense of factor analysis: the use of factor analysis for instrument development in health care research. Sage Publications, 2004.
2) If we are working in the test of theory context, we already know what variables represent (form or reflect) each construct from previous research. For this reason we could use the variables directly (with help of the theory, previous research and your knowledge about the LV and relations).
I hope it helps...
Best regards
Prof. Dr. Diogenes de Souza Bido