"Manifest variable scores (used)" in the SmartPLS

[Lanfranchi]

Hi,


What are the manifest variable scores (used) in the report ? If they are the standardized score, there is a problem. I don't find the same results with Statistica.

I have taken a small example with no missing data (10 subjects and 6 indicators). Statistica and Smartpls use the same mean but not the same SD to calculate the standardized score...

Thanks for you help.

[jmbecker]

First of all YES, the manifest variable scores (used) in the report are the standardized scores.

As you stated yourself, the calculation of the standard deviation is probably not the same.

You should be aware that you can calculate "the sample (or estimated) standard deviation" or "the standard deviation of an entire population". The difference is only the denominator which is N-1 in the first case or just N in the second.

Therefore you should try to change the SD calculation in Statistica or just recalculate the sample yourself.

[Lanfranchi]

Thanks for you answer. But i have already used the SD with N-1 and i don't find the same results !

My data are (from one indicator in "Manifest Variable Scores (Original)") :
4,000000
3,000000
3,000000
3,000000
3,000000
2,000000
3,000000
2,000000
2,000000
3,000000
2,000000
2,000000
1,000000
2,000000
1,000000
3,000000
1,000000
3,000000
2,000000
2,000000

Manifest Variable Scores (Used) :

2,082942
0,820553
0,820553
0,820553
0,820553
-0,441836
0,820553
-0,441836
-0,441836
0,820553
-0,441836
-0,441836
-1,704225
-0,441836
-1,704225
0,820553
-1,704225
0,820553
-0,441836
-0,441836

If i calculate the score with mean = 2,35 and corrected SD= 0,8338, i don't find these scores... I find for the first subject = 1,9789... BUT if i multiply this value by 20/19 (new corrected SD), i find 2,083 !

The standardized scores are calculated with corrected SD and after weighted again ? Why ?

Thank you for your help.

[jmbecker]

I get SD (with N): 0,792148976 and SD (with N-1): 0,812727701 for your example.
If you use the first one you get exactly the same values as in the report.

[Lanfranchi]

O.K. , you win ! Statistica calculated automatically the SD with (N-1) and i believed it was the SD (N)... It standardizes scores with SD (N-1) too.

Normally, i use SPSS which gives automatically SD (N).

Thanks.