Jon Elhai posted on Friday, June 20, 2008 - 9:06 am
Is there a way for Mplus to provide descriptive statistics for categorical variables? Specifically, I'm wondering if there's a way to provide basic frequencies (percentage endorsements of the outcome categories) and cross-tabs of frequencies between two variables.
With TYPE=COMPLEX or TYPE=GENERAL, sample statistics are affected only by weights not stratification and clustering. Those affect standard errors only. You can use the WEIGHT option to obtain sample statistics.
Leslie Roos posted on Sunday, August 18, 2013 - 10:32 am
I wanted to ask a follow-up clarification question regarding descriptive statistics with weighted data. I'm trying to obtain weighted crosstabs percentages based on a complex dataset. I've included the weight, stratification and cluster variable in the input, but the output of crosstabs and frequencies does not appear to be weighted.
TECH10 includes univariate and bivariate frequency information that is weighted.
Leslie Roos posted on Tuesday, August 20, 2013 - 3:33 pm
Thank you so much for that description of TECH10. Although I can see frequency information that is weighted, the crosstabs appear to be unweighted -- Is there any way to have weighted crosstabs (or crosstab frequencies?)
Tamara Fitz posted on Monday, February 15, 2016 - 4:27 am
Dear Prof. Muthen, When I save the data as a .dat type in SPSS and try to run the basic analysis with it, Mplus shows me the warning that there is a invalid symbol in case 1 in the data. As described here i open the data mith Mplus Editor, delete the symbol and save it. Basic analysis is possible now. But when i compare the descriptive results to that from SPSS there are two scales that have exchanged Means (SPSS AB:5,6 CD:3,5; Mplus AB:3,5 CD:5,6) all the other means are identical with SPSS. I dont know how to solve the problem. What could have been gone wrong?
Then try a Model command using WITH for all variables.
Eric M. posted on Monday, December 18, 2017 - 9:49 am
I am attempting to compare the descriptive statistics produced in MPLUS to SPSS output in order to confirm that I have input the data correctly. The two seem off on few values at first glance (In MPLUS I am looking at the section UNIVARIATE SAMPLE STATISTICS). In the MPLUS output I am also getting the means in the section ESTIMATED SAMPLE STATISTICS. I have run syntax using the Type Basic command with sampstat as an output command
1. The means are different within MPLUS sections ESTIMATED SAMPLE STATISTICS and UNIVARIATE SAMPLE STATISTICS, why is that? 2. The means in UNIVARIATE SAMPLE STATISTICS match SPSS output, however variance, skewness, and kurtosis do not. 3. As I attempted to figure out why the descriptives in UNIVARIATE SAMPLE STATISTICS were different than SPSS output, I manually calculated, variance and skewness using excel (using population formula and sample formula). From what I can tell, in the UNIVARIATE SAMPLE STATISTICS section it seems as though SPSS is computing sample variance (as well as skew) and MPLUS is computing population variance in the UNIVARIATE SAMPLE STATISTICS section. Is this correct? If so, although differences are small, which should be reported?
1. There is a difference in estimating means and variances using multivariate information and using univariate information. This is due to missing data and is discussed on our website under Topic 4 and also in our Regression and Mediation book.
2. Skewness and kurtosis exist in two different versions, making them zero or not for normal variables.
3. Mplus does not present the population variance in the Univ Samp Stat section. Mplus uses a division by N for univariate variances whereas N-1 is sometimes used in other software.
Dear Prof. Muthen, Being new to MPlus I'm exploring my dataset first. Running an ANALYSIS: type = basic gives strange results. a) the sample size is off b) the min-max, and thus the mean is off for all my variables.
Some variables have for example a range from 0-52 others 0-6 whereas all should have a range of 1-6. What should I do?
Check your data set for blanks and also that the number of variables in the NAMES statement agrees with the number of columns in the data. If that doesn't help, send your input, output, and data to Support along with your license number.
Skewness and kurtosis can be represented with various statistics. Which of these statistics are computed by Mplus and labeled skewness and kurtosis in the output? I don't see an explanation in the user guide.