Message/Author 

Katerina Gk posted on Saturday, October 12, 2013  10:26 am



Hi I have a two level model. Some variables of my data are answered from teachers and they refer to ther principal. So I decided that those answers I must put on the between level. Do I need to specify the between level I mean BETWEEN IS l1 l2 l3.....or this is by default?. My data include different answers in the same school. Thank you very much for your help 


Be definition, the values of a betweenlevel variable are the same for each cluster member. 

Katerina Gk posted on Monday, October 14, 2013  10:43 am



Dear Linda, Thank you for your reply, make sense what you mean, but the data of the leadership is on the school level. Is it possible to reccomend me something to do in order to use these data only to between part??? I am a little be confused...Next I add my code, my problem is with the variable on the between level l1 l2 l3 l4.... USEVARIABLES = sxoleio l2 l3 ...; CLUSTER IS school; BETWEEN is e1_mm e3_mm e4_mm...; DEFINE: e1_mm= CLUSTER_MEAN (e1); e3_mm= CLUSTER_MEAN (e3); .... ANALYSIS: TYPE IS TWOLEVEL; ESTIMATOR = WLSMV; MODEL: %within% .... %between% er1_b by e1_mm@1 e4_mm...; er2_b by...; er3_b by ...; er4_b by ...; er5_b by ...; par1_b by l2@1 l3 l4; par2_b by l5@1 l6 ; par3_b by l8@1 l9 l11 l12; par4_b by l13@1 l14 l15 l16 l17; par5_b by l19@1 l20; par6_b by l21@1 l22 l23; par_b by par1_b par2_b par3_b par4_b par5_b par6_b; Thank you very much 


If l1 and the other variables do not have the same values for each cluster member, you can create cluster mean variables in DEFINE like you did for the other variables. 

Katerina Gk posted on Monday, October 21, 2013  12:31 am



Dear Linda, Thank you for your message, I have another question I rerun my model(adding the clustre mean for l1...) many times with different combinations trying to fix some of my errors. When I run in the between part the model the second order factor model par1_b by l2@1...; par2_b by l5@1... ; par3_b by l8@1...; par4_b by l13@1...; par5_b by l19@1 l20; par6_b by l21@1...; par_b by par1_b par2_b par3_b par4_b par5_b par6_b; without putting anything in the within level, I dont have any problem. BUT when I add two other models in the within part then the loadings of par3_b and par4_b is greater than one and the residuals vriances are negative but not significant. All the effects I am looking are significant er1_b ON par_b; er2_b ON par_b; er3_b ON par_b; er4_b ON par_b; er5_b ON par_b; As I know it is important that the loadings must not be greater than one, but should I fix this loadings par3_b@0 or what else do you recommend me to do? Thanks in advance Katerina 


Small insignificant negative residual variances can be fixed at zero. Regression coefficients can be greater than one. 

Katerina Gk posted on Monday, October 21, 2013  10:17 am



Thank you very much for your message! 

Katerina Gk posted on Tuesday, October 29, 2013  2:30 am



Dear Linda, I have two queries: 1)is it possible to run and accept the standards error of a twolevel model having only models in between level keeping the within part clear but having in the model? 2)I get negative factor loadings at between level and positive at within and when I test impacts I get negative sign even in within part. Could a possible explanation be that ,between schools, teachers are dissatisfied with their job? My model is measured teachers' job satisfaction. Is this possible or do I make any mistake in the command? WITHIN IS e1...; BETWEEN IS e1_mm...; Missing are all (999); DEFINE:e1_mm= CLUSTER_MEAN (e1); ... ANAL: TYPE IS twolevel; ESTIMATOR IS WLSMV; ALGORITHM=INTEGRATION; INTEGRATION=MONTECARLO; model: %within% er1_w by e1* ..; er2_w by ..; ... er1_wer4_w@1; er1_w ON le; a1_w by a3* a6 ... ; a2_w ... ; a1_wa2_w@1; er1_w ON a1_w; %between% er1_b by e1_mm* ..; er2_b by ..; ... er1_ber4@1; er4_b WITH er5_b@0; er1_b ON lev_mm; a1_b by a3_mm* .. a2_b by ... a1_ba2_b@1; er1_b ON a1_b; ... 


Please send relevant outputs to Support. 

Back to top 