Sample Statistics Check
Message/Author
 Zen Goh posted on Wednesday, June 25, 2014 - 1:59 am
Hi,

I'm running a multilevel moderated mediation (1, (1,1), 1) and am checking my sample statistics with SPSS descriptives to ensure my data is read right.

My output says the following:

ESTIMATED SAMPLE STATISTICS FOR WITHIN

Means
all variables = 0

ESTIMATED SAMPLE STATISTICS FOR BETWEEN

Means
W = 0.004
M1 = 2.483
M2 = 2.321
Y = 3.370
X = 0.000

Covariances and Correlations
- 0 for X with all variables
- other values are close to SPSS's output

Question
1. Is it normal to get mean = 0 for within?
2. Why am I getting mean = 0 for between for only W and X -- these values do not correspond to SPSS's output.

Thank you!
 Linda K. Muthen posted on Wednesday, June 25, 2014 - 2:44 pm
Means are reported on between in multilevel modeling. I doubt that SPSS gives both within and between sample statistics. To compare the programs use TYPE=GENERAL.
 Zen Goh posted on Thursday, June 26, 2014 - 7:57 pm
Hi, thanks for the quick reply!

Yes, SPSS doesn't give within sample statistics- I compared only for between sample statistics. I wasn't clear. I also used your suggestion.

it says that:

Multilevel moderated mediation analysis
L1: X (wload), Y (lsat), M1 (wfc-timestrain), M2 (wfc-emo3)
L2: W (wfc specific-mgsup) - L2
DATA: FILE IS IC_Aggregate_WFCemo and WFCtsb.csv;
VARIABLE:
NAMES = clust x m1 m2 y w;
MISSING ARE ALL (-1);
USEVARIABLES ARE x m1 m2 y w;
ANALYSIS: TYPE = GENERAL;

*** WARNING in MODEL command
All variables are uncorrelated with all other variables in the model.
Check that this is what is intended.
1 WARNING(S) FOUND IN THE INPUT INSTRUCTIONS

2 problems I found:
> The variables are correlated in SPSS but not in Mplus
> my Mplus estimates for W are way too high: mean = 6.2 (it's a 5 point scale)and variance is 312.

Could it be some problem with the way I saved/formatted the data? I simply saved SPSS as a .csv file.

Thank you.
 Linda K. Muthen posted on Friday, June 27, 2014 - 6:25 am
You may be reading your data incorrectly. I would do a TYPE=BASIC with no MODEL command and compare that to the same in SPSS. I would not complicate things with the multilevel component until I was sure that data are being read correctly. Check the means and sample size. You may have blanks in the data set. SPSS uses blanks as certain types of missing data and this is not allowed with free format data.
 Zen Goh posted on Monday, June 30, 2014 - 8:25 pm
It seems to have read it right - the output with what you suggested is very close to SPSS output.

thanks!
 Linda K. Muthen posted on Tuesday, July 01, 2014 - 6:13 am
Then your multilevel results should be correct. I don't believe SPSS does multilevel analysis so the comparison should not be with multilevel results.
 Tyler Burch posted on Friday, October 30, 2015 - 9:08 am
I have a related issue. I run ANALYSIS: TYPE=BASIC and compare it to the descriptives produced in SPSS. The ESTIMATED SAMPLE STATISTICS and the UNIVARIATE HIGHER-ORDER MOMENT DESCRIPTIVE STATISTICS means do not match within the MPLUS output. I am not sure why this is. The UNIVARIATE HIGHER-ORDER MOMENT DESCRIPTIVE STATISTICS descriptive statistics match the SPSS descriptives output perfectly, but I want to make sure that I am not missing something being that the means within the output do not match.
 Linda K. Muthen posted on Friday, October 30, 2015 - 11:10 am
The univariate in Mplus use only cases without missing data one variable at a time like SPSS does. The other sample statistics are estimated in a multivariate fashion using all available data.
 Ashley Hum posted on Monday, October 21, 2019 - 10:27 am
Hello,
I am using type=complex and weighting, stratification, and cluster procedures. Is there a way to get quartiles for analysis variables when weighting is being used?

Thank you.
Ashley
 Tihomir Asparouhov posted on Tuesday, October 22, 2019 - 9:40 am
You can use some code like that to get the estimated quartiles.

MODEL: [y] (m); y (v);
model constraints:
new(q1); q1= m-0.6745*v;
new(q2); q2= m+0.6745*v;

If you want the sample values you can get these in excel by ordering the data and adding up the weights - form a column that has sum(w1:wi) and then find the values in the summed weights column that matches 0.25*N, 0.5*N, 0.75*N
You can use the standardized weights that Mplus produces with the savedata command.
Post:
This is a private posting area. Only registered users and moderators may post messages here.