Message/Author 

Jinxin ZHU posted on Thursday, July 04, 2013  9:12 pm



For ICCS data, if I using multilevel model, there should be two weight variables for students and schools, I should use the option of WEIGHT(student weight information within school, saym, WST) and BWEIGHT (school weight information, say, WSC) in Mplus, right? Thus, I should also specify the CLUSTER for each level, right? To take the sampling error into consideration, I can use Multilevel model, so that I don't have to using the resampling method, right? The option of REPSE do not work in Type=TwoLevel,right? When using Type=Complex, I should firstly generate the replicate weights, using the option of WEIGHT to specify the total weight (WST*WSC, in Q1) of the students, and CLUSTER to specify the school ID the students belonging to, right? There is another option STRATIFICATION, is it used to specify country the schools belong to? When generate the replicate weight, I don't have to specify the MODEL, right? When I use the option of REPSE = JACKKNIFE to counting the sampling error, I should generate the replicate weight using REPSE = JACKKNIFE first, and I should also specify the number of the draws, right? How should I determine the number, if the students are nested in the school and the school in country? Should I using Multi Group analysis to counting the country effect and the option of CLUSTER to counting the school effect within country? 


You must specify the CLUSTER option with multilevel modeling. You can use both a within and a between weight. Replicate weights contain information about the clustering and stratification in the data so the CLUSTER and STRATIFICATION options cannot be used with replicate weights. If you don't use replicate weights, you can use the CLUSTER and STRATIFICATION options. Example 13.9 shows how to generate replicate weights. 

Jinxin ZHU posted on Tuesday, July 30, 2013  12:45 am



According to the User's Guide of MPLUS. "A multiplier file is required for JACKKNIFE when replicate weights are used. The size of this file is one column with rows equal to the number of PSU¡¯s. For each PSU in a stratum, the value in the file is equal to the number of PSU¡¯s in the stratum minus one divided by the number of PSU¡¯s in the stratum. All PSU¡¯s in a stratum have the same value. If replicate weights are generated using JACKKNIFE, a multiplier file can be saved. " Q: Does it mean that we have to generate a multiplier file first when using the method of JACKKNIFE 


Yes, you must create the file. 

Jinxin ZHU posted on Tuesday, July 30, 2013  5:58 pm



According to Muthem' paper "Resampling Methods in Mplus for Complex Survey Data" by Tihomir Asparouhov and Bengt Muth¨¦n, May 4, 2010. "Following is the input file for estimating this SEM model with the 90 replicate weights provided with the data for use with the Jackknife2 method. DATA: FILE IS ECLS repw.dat; VARIABLE: NAMES=C1RGSCAL C1RMSCAL C1RRSCAL P1NUMPLA C1CPTW0 C1CPTSTR NEWPSU W1W90; USEVAR=C1RGSCAL C1RMSCAL C1RRSCAL P1NUMPLA; WEIGHT=C1CPTW0; REPWEIGHT=W1W90; ANALYSIS: TYPE=COMPLEX; REPSE=JACKKNIFE2;" Q1: Does it means that we have to generate the W1W90 ourself£¿ Q2: Can Mplus help generate this weights for JACKKNIFE2? 


Do you want to generate weights or do you have the weights and just want to use them? 

Jinxin ZHU posted on Wednesday, July 31, 2013  6:21 pm



I want generate the weights for my data. For the method of JACKKNIFE when employing the replicate weights, Mplus can help generate the weights. Q3: Can Mplus help generate the weights for the method of JACKKNIFE2?? For example, I change the example 13.19 like this? DATA: FILE IS ex13.19.dat; VARIABLE: NAMES ARE y1y4 weight strat psu; WEIGHT = weight; STRATIFICATION = strat; CLUSTER = psu; ANALYSIS: TYPE = COMPLEX; REPSE = JACKKNIFE2; JACKKNIFE2 = 50;!If the number of cluster is 100? SAVEDATA: FILE IS rweights.sav; SAVE = REPWEIGHTS; 


Following is the input to generate JACKKNIFE2 replicate weights: DATA: FILE=1.DAT; VARIABLE: NAMES ARE weight X psu strat; WEIGHT = weight; STRATIFICATION = strat; CLUSTER = psu; ANALYSIS: TYPE = COMPLEX; REPSE = JACKKNIFE2; MODEL: X; SAVEDATA: FILE IS rweights.sav; SAVE = REPWEIGHTS; 

Jinxin ZHU posted on Thursday, August 01, 2013  7:00 pm



So if VARIABLE: NAMES ARE weight X1 X2 Y psu strat; Then for the option of MODEL: Can I specified as: Model:X1 X2 Y; or any one of the listed are okay? Model: X1; or Model: X2; or Model: Y; 


You should specify the model you are interested in, for example, MODEL: y ON x1 x2; 

Cyndy Karras posted on Wednesday, September 10, 2014  4:01 pm



Dr. Muthen, I am using a large dataset that already contains the basic weight and replicate weights. When I specify: weight = c4natwt; repweight = c4natwt_rep1  c4natwt_rep33; analysis: type=complex; REPSE = JACKKNIFE; model: Y1M_PTNSTR ON Y3M_PTNSTR Y1F_PTNSTR Y3F_PTNSTR; Y1F_PTNSTR ON Y3F_PTNSTR Y1M_PTNSTR Y3M_PTNSTR; I receive this message: "*** ERROR in ANALYSIS command REPSE=JACKKNIFE with replicate weights requires that the MULTIPLIER option be specifed." I have been trying to research what the multiplier option signifies and why I need to use it. I am not sure if I need to create a file since the weights and repweights were already generated. Any insight would be very helpful. Thank you! 


See page 613 of the user's guide where the multiplier file is discussed. 


Hello, I try to employ the replicate weights when analyzing TIMSS 2011 data. I've generated the 75 replicate weights in the data set. However, the linear regression is very strange. These two variables should have a .45 correlation, but the result from Mplus 7.11 shows "no correlation". Did I make any mistake? =================================== TITLE: a linear regression DATA: FILE IS D:\bsgtwnm5_JRR.dat; VARIABLE: NAMES ARE i01i12 TOTWGT HOUWGT SENWGT JKZONE JKREP PVm1PVm5 BSBGSLM RW1RW75; USEVARIABLES ARE PVm1 BSBGSLM; WEIGHT = TOTWGT; REPWEIGHTS = RW1RW75; ANALYSIS: TYPE = COMPLEX; REPSE = JACKKNIFE2; MODEL: PVm1 ON BSBGSLM; OUTPUT: STDYX; 


Please send the output, data set, and your license number to support@statmodel.com. 

Back to top 