|
|
Multiple imputation and small sample ... |
|
Message/Author |
|
Jihyun Yoon posted on Wednesday, December 25, 2013 - 2:34 am
|
|
|
I am reviewing a dissertation having used Mplus for path analysis. The path model the candidate estimated has the information following. Is it ok to conduct MI with such many missing data? With this small sample size (i.e. small power), CFI and RMSEA values are meaningful in Mplus? How DF comes out as 198 in Mplus? -Sample size=93 (The data of 40 subjects were complete for 15 variables, but the remaining 53 subjects had the data for only 7 variables, so the all the data for 8 variables of the 53 subjects were generated using multiple imputation for analysis.) -The number of observed variables in the model=15 -The number of paths in the model=23 Results: CFI=1, RMSEA=0, The chi-square of the model=146.5, DF=198. Thanks! |
|
|
The degrees of freedom in path analysis are determined by the difference between the free parameters in the H1 and H0 models. It would require a study to see if imputation would work in your case. |
|
Jihyun Yoon posted on Friday, December 27, 2013 - 12:34 am
|
|
|
cfi=1 and rmsea=0 seemed to come out because df is bigger than chi- squred value. So in this case, we can not use cfi or rmsea as model diagnostics statistics, can't we? |
|
|
Please send the output and your license number to support@statmodel.com so I can see what you are looking at. |
|
Back to top |
|
|