- Видео 34
- Просмотров 36 656
benR
США
Добавлен 24 фев 2021
Factorial ANOVA in R is Quite Easy
This video walks through how use R to run a factorial ANOVA (simple 2x2 between subjects example), get effect sizes, and look at pairwise comparisons. It's easy : )
Просмотров: 134
Видео
Data Cleaning Part 3
Просмотров 397 месяцев назад
This video shows how to compute a time taken on study variable using Qualtrics' duration in seconds variable and use it to filer out too fast or too slow completions.
Data Cleaning Part 2
Просмотров 477 месяцев назад
This video shows you how to identify rows with incomplete data and remove them from your dataset
Data Cleaning Part 1
Просмотров 1257 месяцев назад
This video shows you how to identify bad responses and remove them from your dataset.
Multiple Group Moderation
Просмотров 926Год назад
This video demonstrates how to use structural equation modeling (SEM) to test moderation when your moderator is a grouping variable. For example, if you conduct a study in two different samples, SEM can help you demonstrate measurement invariance across the samples and then determine whether the effects under examination are the same in both samples.
Import Data from Qualtrics - EASY!
Просмотров 439Год назад
Data directly downloaded from Qualtrics can be obnoxious to import in R because it always adds an extra row. Fortunately, the qualtRics package provides an easy solution to the problem.
Help! CFA Model Fit Part 4: Parceling
Просмотров 902Год назад
This video demonstrates how parceling can improve model fit. It is Part 4 of a four-part series on problems with model fit in confirmatory factor analysis for structural equation modeling.
Help! CFA Model Fit Part 3, Modification Indices
Просмотров 1,5 тыс.Год назад
This video demonstrates how to generate and interpret modification indices for a CFA, and how to use this information to diagnose potential problems with model fit.
Help! CFA Model Fit Part 2: Residual Matrix
Просмотров 790Год назад
This video demonstrates how to generate and explore a residual matrix using Lavaan for R. I demonstrate how to interpret the residual matrix to understand areas of strain in the model and diagnose why the model fit might be poor.
Methods Sections
Просмотров 166Год назад
This video is designed to walk people through the steps needed to prepare the methods section of a manuscript. It uses the tabyl function of the janitor package to get frequency statistics, the describe and summarize functions of psych and dplyr to get means and standard deviations, and the cronbach function of psy to get Cronbach's alpha. It also illustrates reverse coding and dummy coding usi...
Missing Data Part 2
Просмотров 48Год назад
This is the second video of three in a missing data series. It shows how to use mutate_at to replace missing values with the median and how to use mutate_at to replace missing values with a zero.
Missing Data Part 1
Просмотров 77Год назад
This video shows how to inspect data to identify missing values, how to create a count variable for each observation summarizing the number of missing columns for each row, and how to remove rows with a determined number of missing cases.
Recode Multiple Items Using Mutate At
Просмотров 1,8 тыс.Год назад
This video illustrates how to recode (reverse code) variables in R using dplyr with the mutate_at command. This illustrates how to recode multiple items at once, though users can do one item at a time using mutate.
Missing Data: Replace Codes with NA using NA_If
Просмотров 815Год назад
Missing Data: Replace Codes with NA using NA_If
Multi-Group CFA: Measurement Invariance
Просмотров 3 тыс.3 года назад
Multi-Group CFA: Measurement Invariance
Mediation in Structural Equation Modeling (SEM) Using Lavaan for R
Просмотров 14 тыс.3 года назад
Mediation in Structural Equation Modeling (SEM) Using Lavaan for R
UnboxingR: Introduction to R and R Studio for statistics
Просмотров 1883 года назад
UnboxingR: Introduction to R and R Studio for statistics
nice
Fantastic explanation. Thank you very much! I want to ask how to proceed when comparing indirect effects between the two countries? I assume that for specifying an indirect effect one would write: indirect := auth2ft * ft2bi (notation X2Y means path from X to Y) Now, where to put the list command (c)? In front of the a-path (auth2ft)? In front of the b-path (ft2bi)? In front of both paths? How to proceed then!?
Thank you so much!
Thank you so much for this video, it was very helpful! One quick question. Are the individual variables allowed to co-vary?
Sir, I have analysed 7 fraction s of soil zinc & different soil parameters of paddy field. How can I develop a SEM including parameters that I have analysed and the yield of paddy? Could your please share your knowledge & teach me.
Hi Ben! Thank you for the video! I did a three-wave study, repeating the measures in each wave. Does it make sense to use the adjust my CFA if, for instance, I get high MI between same itens in different points in time, or if I get high MI between different itens in same points in time. Example of one and other: Time1_Bullying1 ~~ Time2_Bullying1 (1 correspond to the item number) Time1_Bullying1 ~~ Time1_Bullying2 (1 and 2 corresponde to item number) Thank you so much!
You're the best. This just saved me the hours I would've wasted hard coding this
this one is good but can i know how we can change legend name in path diagram's boxes
very much helpful. can you please share your email address sir
Excelent video, Ben! A question, even though the indirect effect is significant, would the Polarisation's small R2 be a problem?
Excelent. Finally I understand about missing values using R
I do not know what is going wrong in my dataset, but I keep getting the following error. Any suggestions would be much appreciated. Error in lavaan::lavaan(model = model1, data = mydata_dmc, group = "coping", : lavaan ERROR: mismatch between number of groups in data, and number of groups in model.
Can you do one with an interaction between an observed variable and a latent varaible! <3
good
Hi ben! After we get the average of the factor loadings for each parcel, how do we use it? What I mean is that do we put is it our data and each respondent will have the same parcel value? Or do we use it as the fixed value of the parcel?
Could we test multiple indirect effects in a single model? Like using ind1, ind2, ind3...?
thanks!
Dear Ben, Thank you very much for this amazing explanation of Mediation in SEM. I have a question, if I may. I am running a mediation model in SEM with two continuous predictors, three continuous moderators, one continuous outcome. I let the predictors co-vary with each other, and the moderators co-vary with each other. I want to compare the strength of the indirect effects of these predictors on the outcome via the parallel mediators, but I do not know how to conduct this comparison in R. Most example mediation models I have seen so far only work with one predictor. Even if they have multiple predictors, they do not compare the indirect effects between predictors. I would appreciate if you could give me some leads, or publish a video on this topic. Thank you very much for your time, and I look forward to hearing from you soon.
Hi Ben, this was really useful, thank you! Are you still planning on making a follow-up video about data cleaning?
Yes, thank you! (I knew I was forgetting something : )
Hi. how would the indirect effects be written if i had independent variables and control variables? so like a1 + a2 =..
Thank you for creating this helpful video!
Thank you so much. Can you also do a video illustrating how to select the best model?
Thanks Ben
Thank you for putting this video together! It's well demonstrated and helpful to me.
Thank you! I can't tell you how helpful all of your videos are! I just wanted to ask if it makes more sense to go through the equality constraints with a fully adjusted model (i.e. including covariates) or with just the DV ~ IV pathways? I suppose I'm asking where you'd recommend placing covariates in this process? Thanks again! 😊
A good question. I suppose this is a matter of taste. Lately, I think I prefer to have my "final model" relatively clean (i.e., only including theory-suggested variables) and the model with a standard set of covariates run as a robustness check on my final model. So, ideally, I'd like to leave the covariates out of it and report an appendix that shows my effects are unchanged when covariates are included. Of course, if the covariates do change the result, simply proceed as directed by theory. As to the when / I'd probably like to get the theoretical model fit first, then add the covariates at the end, just because sometimes adding in a bunch of observed variables into the regression section of the model can cause some unpredictable things to happen and it is easier to diagnose what's going on when you understand the base model first. This is my advice in general, but especially valuable for a multiple group model with a bunch of equality constraints because it introduces so many more opportunities for a theoretically trivial area of misspecification .
@@benr1813 That makes a lot of sense, thank you!
@@benr1813 Hi again! I had a bit of a follow-up question - would it make sense to apply a correction for multiple tests when comparing the chi squared outputs, or should each path comparison just be interpreted on it's own? Thanks again!
@@tomoskipididoo Good question, it kinda depends on your a priori expectations I suppose. Rex Kline, who is influential in SEM across the social sciences, doesn't like the hypothesis testing approach to SEM and would probably advise you to decide if your most parsimonious model causes you to accept or refine the theoretical model you established a priori. I personally don't know that the global model decision is a tenable approach to evaluating each path and have found that readers and reviewers prefer to evaluate each path one-by-one. Thus, I view each path as an individual hypothesis and test it without a multiple test correction, with the logic that the directional hypothesis for each path means I'm only conducting one test for each component of the theoretical model. (aside, I really like Andrew Hayes writeup on the multiple test problem in his book Statistical Methods for Communication Science). However, especially if your model is exploratory or your theoretical justification for paths is tentative, I can see wanting to do a correction for multiple tests. Alternatively, any modifications you make that are based on data and retrospective sense-making should be considered provisional until replicated in fresh data. So my approach is more: theoretical hypothesis = path, modification = post-hoc and provisional subject to revision.
@@benr1813 This is such a brilliant reply, I really appreciate your help. Each path being an individual hypothesis test makes sense, and I'll be sure to read the Hayes write-up too! Thanks again! 😊
Very interesting.
What does the "rsquare" mean in the model output? Thanks.
It means "proportion of variance explained." For the indicators of latent variables, it means amount of information an indicator is contributing to the latent variable. For latent variables, it means how much information is being explained by the independent variables in the model.
@@benr1813 Thank you very much for the explanation! I wonder why the "rsquare" output does not include "VCneg"? Also, the "rsquare" of "Pol" is only 0.034, does this mean that the model is not able to capture the variation of "Pol" as only 3.4% has been explained by the model? Thank you.
Thank you for the great work! Is it possible to share the dataset in the example?
Thanks Ben. Looking forward to the next video.
Thanks for this whole series, Ben.
To increase Max Print, use options(max.print = 10000)
I found this video very helpful. Thank you
You know your stuff. Thank you so much. This has been really helpful
Thanks it is very help
I am getting an error "If the match-paired approach is used, the number of variables in all sets must be equal." I have unequal no of items for those variables ! Also i wish to know how to handle multiple moderators acting on single IV
So helpful!! Thankyou
Thanks for your video! It was very useful! I have a question. At 2:10 , why did you choose the CFI coefficient on the left instead of the one on the right or the Robust CFI? Since you used maximum likelihood estimation with robust standard errors (MLR), I thought you would use one of those two. I'd glad if you could clarify that for me
Thank you, Ben! Finally I found a video to solve na values!
This was really helpful. Thank you!!!!!!!!!!
This is one of greatest tutorial I've ever seen!
Hi Ben, thank you for sharing this video. I am doing a GDP and stock price correlation in 5-years currently. I am struggling how to create data in SPSS. since each year only have 1 GDP number, but many number of stock price (50 stocks). Could you help me on this?
I am grateful to you teacher!
Thanks i was battling with a recode this worked perfectly
Hello, Is that possible to do path analysis with the economic panel data series? I mean, with a reasonably large panel ... Thank you a lot.
Thanks so much for this tutorial! Straightforward and easy to follow instructions. Have you made the follow-up video without the matching?
that was very useful for me. thank you very much
Can you please show us how to compare the two latent means after establishing the scalar invariance?
This is a good idea for a video! I'm not sure how soon I'll be able to record one but here is some guidance whilst I get it together: (1) if you use fixed-factor method of identification then one group's mean will be fixed to zero and the estimate of the other group's mean will be the difference between the two groups. You can then fix that value to zero and use a chi-square difference test to see if the difference is statistically significant. (2) you can use Todd Little's effects coding method to generate latent means that are interpretable in the metric of measurement and create an equality constraint for the chi-square difference test. The two approaches are the same but the effects coding method gives you latent means to report.
Thanks for the video. I see that the IV and the moderator all had 3 items, is it necessary for them to have same number of items? I if have IV with 3 items and moderator with 5 items, should I parcel? thanks a lot.
Hi Yuanjie, it is not necessary that the IV and M have the same number of items but it gets complicated fast so I do recommend parceling.
Hi, I have a question about the model. Is Negative Contact a Szenario Variable/ dichotomy Variable or binary?
Hi! can you explain how to deal with control variables? Have they to be addedd in the model as well? thanks!