Hi all,
i wanted to test my data for normality and need to confim which one of the ways is the right one:
1. do i conduct normality test for each "Item" i.e indicator of the constructs, or
2. should I do it for each construct i.e. Latent Variable in my Model for e.g. by importing the LV unstandardised scores from Smartpls output in SPSS
Any help would be greatly appreciated.
Thx in advance.
Test of normality
-
- PLS Junior User
- Posts: 4
- Joined: Fri Oct 30, 2009 5:15 pm
- Real name and title:
-
- PLS Junior User
- Posts: 4
- Joined: Fri Oct 30, 2009 5:15 pm
- Real name and title:
-
- PLS Expert User
- Posts: 248
- Joined: Sat Jul 25, 2009 1:34 pm
- Real name and title:
Hey Kunal,
Some hints, that maybe will help you:
ad 1) Generally, you have to test each and every item (indicator) if the follow a normal distribution. I have often seen that procedure in articles.
ad 2) I would test the indicators unstandardised.
Furthermore, they typically tests which are used (e.g. Kolmogorov-Smirnoff-Test) are to conservative for quasimetric scales. Much better for testing the distribution of your indicators are visual inspections, such as QQ-plots or you look at the ‘kurtosis’ and ‘skewness’. Is the absolute value of the ‘kurtosis’ and ‘skewness’ higher than 1, you can say, your indicator isn’t normal distributed. SPSS is very useful for conduct such a normality test (a german hint for doing that in SPSS: Analysieren → Deskriptive Statistiken → Häufigkeiten unter dem Bereich „Statistiken“ anklicken).
Kind regards,
Christian
Some hints, that maybe will help you:
ad 1) Generally, you have to test each and every item (indicator) if the follow a normal distribution. I have often seen that procedure in articles.
ad 2) I would test the indicators unstandardised.
Furthermore, they typically tests which are used (e.g. Kolmogorov-Smirnoff-Test) are to conservative for quasimetric scales. Much better for testing the distribution of your indicators are visual inspections, such as QQ-plots or you look at the ‘kurtosis’ and ‘skewness’. Is the absolute value of the ‘kurtosis’ and ‘skewness’ higher than 1, you can say, your indicator isn’t normal distributed. SPSS is very useful for conduct such a normality test (a german hint for doing that in SPSS: Analysieren → Deskriptive Statistiken → Häufigkeiten unter dem Bereich „Statistiken“ anklicken).
Kind regards,
Christian
-
- PLS Junior User
- Posts: 4
- Joined: Fri Oct 30, 2009 5:15 pm
- Real name and title: