Intraclass Correlation in Mplus

[@dadovader]

Very helpful video! I appreciate how you walk through the ICC's importance. Additionally, I appreciate you addressing the type of ICC in the comments (which is originally what I came to this video to find out).

[@QuantFish]

Hi Carlo, Thank you very much for your comments. I'm so glad this was helpful for you. Best, Christian Geiser

[@crazystatistician]

Hi, i am wondering whether can we use the means of variables which is calculated in SPSS when conducting ICC in Mplus using the syntax explained by you? or should we use latent factors where we should describe some items loaded on the corresponding latent factor such as satisfaction BY a1 a2 a3 a4? Which one can we prefer when calculating the ICC? Thank you in advance!

[@alperuslukaya8874]

Thank you very much for the video. I'm conducting a multilevel analysis, and I would like to know how to calculate Rwg values. Do you have any information on this?

[@CightFlub]

This is such a helpful video! Thank you very much!

[@çağla-i5y]

Hello Thank you for the video. I want to ask about the dail-data. I have daily measured data for 5 days from 180 people. Measures have muktiple items, so how can i do ICC analysis in that case?

[@QuantFish]

It should be no problem as long as you have your data in long format with a time variable indicating the days (in case your data is in wide format, you can use the wide-to-long transformation option in Mplus; see my video on that topic). Best, Christian Geiser

[@kabirumaitamakura6388]

This is interesting. I appreciate if you can share the data file so that we can follow.

[@ravenrinas]

This helped me a ton - thanks so much! Is there a particular ICC cut off value that would indicate that nested analyses are required?

[@QuantFish]

Hi Raven, I'm not aware of a particular cutoff but would recommend that you always take the nested data structure into account since even small-looking ICCs can lead to Type-I error inflation (incorrect rejection of the null hypothesis in too many cases). Best, Christian Geiser

[@ravenrinas]

@@QuantFish Thank you!

[@j_5chneider74]

Very nice, thanks! One question: Which kind of ICC is MPlus computing? Shrout and Fleiss (1979) define several ICCs usually referred to as ICC1, ICC2 and ICC3.

[@QuantFish]

Thank you for watching! You can find the formula that Mplus uses to compute the ICC in the following technical appendix (Equation 203): www.statmodel.com/download/techappen.pdf Best, Christian Geiser

[@longzhudong2777]

@@QuantFish Thank you so much! Your videos are always clear and very helpful. A following up question, can Mplus compute ICC2, ICC3, and rwg? Do we need all the ICCs and rwg to justify aggregation?

[@QuantFish]

@@longzhudong2777 Thank you for your positive feedback! I'm not sure about the other ICCs. This question would be best directed towards the Mplus support team. Best, Christian Geiser

[@alperuslukaya8874]

@@longzhudong2777 Did you manage to find any information on how to calculate the Rwg value? Can you help me with that?

[@crazystatistician]

Do the data need to be in long format for this model?

[@QuantFish]

Yes, the data must be in long format for any kind of analysis that uses TYPE = TWOLEVEL in Mplus, and a cluster (grouping) variable needs to be included in the data set to identify Level-1 and Level-2 units. Best, Christian Geiser

[@crazystatistician]

Thank you so much!@@QuantFish

[@lunawong4039]

Hi Dr. Gelser, when I tried to run the model, I got an error message. I tried to use different variables, but still got the same error. Could you please advise what I should do? A MATRIX COULD NOT BE INVERTED DURING THE H1 MODEL ESTIMATION. THE ESTIMATED WITHIN COVARIANCE MATRIX IS NOT POSITIVE DEFINITE AS IT SHOULD BE. COMPUTATION COULD NOT BE COMPLETED. THE VARIANCE OF COMMIT1 APPROACHES 0. FIX THIS VARIANCE AND THE CORRESPONDING COVARIANCES TO 0, DECREASE THE MINIMUM VARIANCE, OR SPECIFY THE VARIABLE AS A BETWEEN VARIABLE. THE H1 MODEL ESTIMATION DID NOT CONVERGE. THE SAMPLE STATISTICS COULD NOT BE COMPUTED.

[@QuantFish]

Have you checked the variance of the variable COMMIT1? It sounds like this variable may not have variance on Level 1 (the WITHIN) level. Is there perhaps an issue with your data file (i.e., does Mplus read it correctly?)? Best, Christian Geiser

[@lunawong4039]

@@QuantFish Hi Dr. Geiser, thanks for the advice. I figured out the error!