Monday, April 12, 2010

A Little Bit of Data Analysis

1) What to use?
2 main alternatives :- a) SPSS
b) SEM- eg: EQS, LISRELL , AMOS, RAMONA, PLS

2) When to use?
a) SPSS - model not that complex
- sample size idak le besar benor.

b) SEM- cater for more complicated model

3) SEM; nak guna yang mana?
Basic idea; PLS is used when - i) sample is small
ii) want to test the theory/ theory not strong
iii) predictive nature of the reserach/developed model
Min sample size requirement for PLS:
10x the number of predictors from (a) or (b), whichever is greater:
(a) the indicators on the most complex formative construct; or
(b) the largest number of antecedents constructs leading to an endogenous construct
(Barclay et al., 1995)
Note: Kalau tak ada formative construct, requirement (b) must be met.


AMOS/LISRELL etc is used when - i) large sample size
ii) theory is strong

Minimum sample size?- 200 pun ok dah to run SEM (Hair et al., 1995).

4) Apa nak buat b4 nak buat analysis?
What ever SEM tools you pilih you must:
a) understand about reflective & formative indicators of your construct coz PLS can run kalau ada kedua dua ni but yang lain rasanya cater untuk satu je.
b) kena la prepare your data dulu ;
i) missing data analysis- check on number & pattern
ii) decide on the treatment of missing data
iii) check outliers
iv) check normality- except kalau guna PLS; tak yah buat benda ni.
v) then baru lah buat preliminary analysis.

Note: to the readers out there, kalau salah cakap ye...saya kan "budak" baru belajar..kalau salah tolong tunjukkan.....

No comments: