วีดีโอ

Hello dr geiser can you upload a invariance analyses for bifactor model in mplus? For categorical modeli

Hi Dr Geiser i have a question when I compare three model to each other configure metric and scalar they are all significant but I have 3000 sample size which means I can ignore this statistics and go on to the CFI TLI and RMSEA so when I go there they are mostly 0.80 in this case as far as I know I do not have to focus on whether they fit percent or not I need to focus on the CFI and Tli differences between metrics if they are below 0.010 and RMSEA 0.015 i hope you understand. Or as Chi square is significant and chi tli are all below 0.90 should I claim that I can not do comparison

Hi I like your content but you have a tendency to smack your lips right into the mic and and it's quite distracting especially with headphones. If you listen back you will hear that you do it every few sentences Just be more mindful of the mouth noises you are making into the mic and/or back off it

Dear Dr. You suggested that we use methods such as Monte Carlo simulations to plan the sample size before the study. Do you have a video on this?

Do I need to show model fitness when I draw simple path analysis between two variables in Amos? And do I also have to add the SEM model in each Hypothesis testing? Unfortunately none of my professors have done any thesis interpretation on Amos so I'm a bit clueless along my professors!

Thanks Christian. Really enjoy your videos. I'd be interested in a video on exploratory SEM, particularly 'CFA-within-SEM'.

I dont get plot from my mplus (ver 8.6). It says that it is not provided. Do you know why that happens? Do I need another version?

Thank you so much, this was very helpful!

Thank you for your excellent videos on CFA/SEM analysis. Please consider adding a video on the different estimation methods (e.g., ML, RML, DWLS, etc) available in JASP and lavaan. When should one consider using each of the estimation methods?

In the data, you have one row per participant, and one column per variable (single-level wide format)?

what would happen if we had also memory3 memory4 memory5 and SWB3 SWB4 etc? do you have video for that?

such a great video! Thanks

Hi Dr. Geiser, This is a great video! Thank you for everything you do! My indicators have 4 response categories, so I'd like to use the WLSMV estimator. Is there a way to do it using this method of Equivalence testing?

Thank you QuantFish!

fantastic explanation!

I don't really understand what LCA and factor analysis have in common

Hi Dr. Geiser! I am wanting to reproduce this MC simulation on my own (very complex) data set to estimate post-hoc power. Most of my observed data is categorical (Likert scale data). The model I ran is a CFA analysis testing longitudinal measurement invariance, using a WLSMV estimator and Theta Parameterization. Here are the 3 problems/ questions I ran into: 1- All my observed variables are categorical. As such, I have added some specifications in the preliminary input (equivalent to your CFA 2 factor instructions). The commands go as such: TITLE --- DATA --- VARIABLES .. Names are.. IDVariable.. Missing.. Usevariables.. Categorigal. -- ANALYSIS .. Estimator (WLSMV).. PARAMETERIZATION (Theta). -- MODEL -- OUTPUT -- SAVEDATA. Is it approrpriate to add the categorical, estimator and parameterization specifications? How would that then be reflected in the MonteCarlo input instructions? 2- In my original CFA input instructions, my "Variables =" and my "usevariable =" are not the same, such that the dataset has more data than is used for the analysis. When I save the model parameters (in your example, est.dat), and then run the montecarlo instructions, I have many warning messages telling me that some variables are uncorrelated with all other variables, which makes sense because they are not the variables under evaluation. Should I only enter in the montecarlo input the variables in my model that were under "usevariable"? Relatedly, I also get error messages that tell me that my COVERAGE file (i.e., est.dat) does not contain enough data. Is that because it expects all the listed variables rather than the actual used variables on which the "est.dat" parameters are based? 2- Is there a possibility (or even need) to specify missing data? for example, in my original CFA input instruction, I could specify " MISSING = ALL (-99). Is it because this is a simulation, and as such, there is no "actual" missing data? Thank you for any help you can provide. Your videos are always helpful. Best!

Conceptualizing measurement invariance as "measurement equivalence" was very helpful for me to understand. Thank you!

Vielen Dank 💐

Vielen Dank für die Videos ! Die Methodenberatung meiner Uni hat mir den Quantfish Kanal empfohlen. Da ich neu im Thema Multigroup- SEM und Messinvarianz bin, haben mich die Videos sehr unterstützt. Nach vielen Recherchen ist für mich immer noch eine Frage offen: Wie passen die Modelle der Messäquivalenz (essenzielle Tau -Äquivalenz und Tau -Äquivalenz) der KTT zu den konfiguralen, metrischen und skalaren, strikten Messinvarainzen ? So wie ich es verstehe, ist die Tau kongenerischen Messäquivalenz mit dem konfiguralen Basis Modell Fit über SEM zu prüfen. Wie kann ich Tau -Äquivalenz prüfen ? Reicht dafür dass das metrische Modell der Messinvarianz angenommen wird mit nicht signifikantem Chi- Quadrat Differenztest oder brauche ich zusätzlich noch die Annahme der skalaren Messinvarianz?

Thanks. Great introduction to LPA

Thanks very much for this video, it is really helpful.

Sir..can we use efa extracted factors to do sem without doing cfa?

Hi and thank you for very helpful videos! I have been puzzling with one issue in terms of multilevel regressions: sometimes it happens that (standardized) regression coefficients have values greater than 1 at the between level (or even values like 3.2 or more). I've learned that this is likely to happen when there are only few variables in the model and when there is only little variance to be explained on the between level. In my example the studied relation is autoregression between the same variable at T1 and T2 (when multicollinearity is obvious). But I would like to know if there is some other explanations for the issue and how this kind of results and models should be interpreted (or should they?). Thank you if you have time for response!

Hello I just saw this tutorial and I would like to know what would be the difference or why would someone would choose to use JASP versus Python? What does JASP do better?

Thank you for the video! What are the next steps once a suppression effect ? How would we write this up in a results section for example?

Hi! If some of my variables correlate with each other, would it make sense to allow them to covary in each latent class (rather than just in one, as in your example)? Thank you!

That is so helpful for my research! thanks a lot

Your video was incredibly helpful. It answered a question I've had for a long time. Thank you for the effort you put into creating this content.

What should we do if the model fit indices do not align with the data, but we still want to create a total score? (Modification indices cannot be used, as the total score will not be impacted by improving the model).

Interesting example, I will keep that in mind. Thanks for the explanation.

I am running an SEM that has a couple of factors with only two indicators. I am struggling whether to use a summated scale or a factor. Your video was a tremendous help in lessening the struggle.

Thank you!

Thank you for this tutorial! I have missing data in my file and receive an MPlus error message that indicates - "TECH13 option not available for TYPE=MIXTURE with missing data." Any suggestions for a workaround?

Hi Dear Dr. Geiser I have a three waves of data and t1 t2 and t3 because t2 is greater than t3 I can not do LGM. But when I free parameter t1dep@0 t2dep* t3 dep@3 it does not give me solution what should I do. and also I have another variable with T1-PPU T2-PPU and T3 PPU and I restrict all of them as there is no problem with ppu T1 ppu@0 t2ppu@1 t3ppu@2 how can I solve this

thank you for your useful work TECHNICAL 13 OUTPUT SKEW AND KURTOSIS TESTS OF MODEL FIT PROBLEM OCCURRED DURING THE COMPUTATION OF MULTIVARIATE SKEW AND KURTOSIS TEST OF FIT. THIS IS MOST LIKELY DUE TO NEARLY SINGULAR INFORMATION MATRIX. I don't understand what I did wrong. any advise?

thankyou

I feel stuck about my data and found this video. Really helpful. Thanks a bunch!

My own personal opinion: R - excellent but requires a lot of practice in coding to become competent in its use. Data visualisation leaves a little to be desired - but its capabilities in data manipulation are second to none. JASP/JAMOVI - If SPSS and R had a baby! Intuitive and easy to use, but lags quite a bit with large and more complex datasets. Still missing a lot of features of that will hopefully be added soon! MPlus - Minor learning curve and quite expensive, but in my opinion the best tool for SEM. Powerful but a little clunky! For coding rookies: 1) Jasp 2) MPlus 3) R For coders: 1) R 2) Jasp 3) MPlus

Dear Christian, Thank you very much for this helpful video. I have a question regarding the number of measurement points needed to model the following scenario: A shock occurs at time t, resulting in a sudden decrease in the variable Y (a drop between t and t1) that was previously stable (e.g., no changes between t-1 and t or t-2 and t-1). After this, individuals may recover, and Y increases again (an increase between t1 and subsequent points, such as t2, t3, t4, etc.). Thank you very much in advance for any tips!

Thanks for your video! Very interesting! So, the suppressor effect is not a problem when we use the model to predict the y variable, right? I mean, in the model included x1 and x2, we can not say the verbal ability has a negative effect on the job_success, right?

Good explanation sir..

Hello Dr. Geiser, Is there a cutoff value for discrimination such as in CFA item loadings (e.g between 0.3 and 1.0)? I have some items having discrimination values above 1.0 and wonder if they mean something is off. Thank you for the awesome video then again!

JASP needs power analysis for the ANOVAs too.

Hi Dr. Geiser, thank you for another informative video. I have conducted LPA in latent gold, and the output provides loadings for each indicator on the variable "cluster." These are provided in addition to the regression parameters. I am wondering what these loadings represent? Are these akin to factor loadings? How do they differ from the regression parameters?

Why could this not be shown on an Y and X axel? With time being the X-axel, slope being the line/curve between the different points, intercept the starting point on the y-axel and epsilon makered by different colors?

Thank you so much Dr. Geiser. I refer to your videos so often that I feel like I should list you as a co-author in my work:) Could you point to some references on using CFA without EFA?

Thank you for the video! Is DIFFTEST also sensitive to large sample size?

Dear Dr. Christian Geiser, your videos are really helpful to me. I have a question regarding the error I got. The idea is that I have a 3-level hierarchical dataset (longitudinal time points nested in family IDs and then nested in site IDs), and I'd like to check how many classes in this data. When I include the covariates (age and sex) into the process to determine the class membership. This is my script: TITLE: "Trajectory modelling of depression scores" DATA: FILE = b4_NOTign.csv; VARIABLE: NAMES = depm0-depm5 deps0-deps5 age4 agepre sex r_white r_black r_other r_asian r_aian r_nhpi r_mixed income under_pov pub fam_id site_id age age_Mar20 rid edu_p; !age is latest age USEVARIABLES = depm0-depm5 age sex fam_id site_id; CATEGORICAL = sex; CLASSES = class (3); CLUSTER = fam_id; STRATIFICATION = site_id; MISSING = ALL (999999); ANALYSIS: TYPE = TWOLEVEL COMPLEX RANDOM MIXTURE; MODEL: %WITHIN% %OVERALL% intw slopew | depm0@0 depm1@1 depm2@2 depm3@3 depm4@4 depm5@5; slopew@0 %BETWEEN% %OVERALL% intb slopeb | depm0@0 depm1@1 depm2@2 depm3@3 depm4@4 depm5@5; slopeb@0 intb slopeb ON age sex; However, I always get this error: *** ERROR in MODEL command Unrestricted x-variables in TWOLEVEL MIXTURE analysis must be specified as either a WITHIN or BETWEEN variable. The following variable cannot exist on both levels: AGE *** ERROR in MODEL command Unrestricted x-variables in TWOLEVEL MIXTURE analysis must be specified as either a WITHIN or BETWEEN variable. The following variable cannot exist on both levels: SEX Could you please help me troubleshoot this? I really appreciate your help!

Thanks for the video. What about when the first part of the output for the configural, metric, and scalar models have significant chi square values but the comparison models (e.g., configural vs metric, metric vs scalar, etc.) do not?