- 40
- 29 623
Caspar van Lissa
เข้าร่วมเมื่อ 7 พ.ค. 2009
GLM: Repeated Measures ANOVA
Repeated Measures ANOVA is used to analyze data collected in within-participants designs, where the same outcome measure is collected from the same individuals multiple times. This lecture discusses the advantages and disadvantages of within-participants designs, their use in different types of research, and the two common ways to analyze repeated measures data: the linear mixed model, and multivariate approach.
มุมมอง: 573
วีดีโอ
GLM: ANCOVA
มุมมอง 21011 หลายเดือนก่อน
ANCOVA, or Analysis of Covariance, is essentially multiple regression with a categorical predictor and one or more continuous predictors. What’s “special” about this technique is that it is commonly used when the continuous predictor(s) are so-called “covariates”: predictors that are only included to improve our estimate of the effect of the categorical predictor of interest. This lecture refre...
Psychometrics II: Dimension Reduction
มุมมอง 16211 หลายเดือนก่อน
This lecture introduces three techniques for reducing multiple items to a smaller number of variables: Principal Components Analysis (PCA), Exploratory Factor Analysis (EFA), and a little bit of Confirmatory Factor Analysis (CFA). Our focus is on PCA and EFA, where PCA is a data rotation technique that can be used for dimension reduction - and EFA is a latent variable model that assumes the exi...
GLM: Factorial ANOVA
มุมมอง 19311 หลายเดือนก่อน
Factorial ANOVA is used to examine the effects of multiple categorical predictors and their interactions on a continuous outcome variable. Factorial ANOVA can be conceptualized as a special case of multiple regression. Each factor is represented via dummy coding, and creating interaction terms are computed by multiplying those dummies. The unique contribution of each factor and the interaction ...
GLM VI+: Logistic Regression
มุมมอง 17011 หลายเดือนก่อน
When our dependent variable is binary (coded as 0 or 1), we can use logistic regression. Instead of predicting the raw scores, we predict a transformation of the dependent variable. Specifically, we model the log odds of the probability of Y being one category (e.g., 1) versus the other category (e.g., 0). This results in an s-shaped regression curve bounded by 0 and 1. Logistic regression assu...
Psychometrics I: Scale Reliability and Validity
มุมมอง 23411 หลายเดือนก่อน
Questionnaires are widely used in the social sciences to measure constructs, including self-reported behavior, beliefs, knowledge, opinions, values, attitudes, and attributes. Researchers may want to know whether these questions do a good job at measuring the construct of interest. Reliability means that the scale measures the same construct repeatedly, and validity means that the scale measure...
GLM VI+: Interaction Effects and Simple Slopes
มุมมอง 377ปีที่แล้ว
An interaction effect occurs when the effect of one predictor depends on the value of another predictor. This lecture introduces interaction effects between continuous and binary variables, and between two continuous variables. It explains how to calculate a product term, discusses the importance of centering, and explains simple slopes analysis for binary and continuous moderators.
GLM VI+: Coding Schemes, Planned Comparisons and Post Hoc Tests
มุมมอง 217ปีที่แล้ว
This lesson delves deeper into categorical predictors in the general linear model or ANOVA, addressing dummy coding, effects coding, and contrast coding, including planned comparisons. Moreover, it addresses the idea of post-hoc tests: comparing the means of all groups with one another, while controlling the experiment-wise Type I error rate.
GLM VI: Nested Models
มุมมอง 802ปีที่แล้ว
Nested models refer to models that are identical, except for the fact that some parameters are constrained in one of them, while all parameters are free in the other. The smaller or constrained model model is said to be “nested in” the larger or unconstrained model. Incremental F-tests are used to determine whether the increase in between two nested models is statistically significant. These te...
GLM V: Multiple Regression
มุมมอง 613ปีที่แล้ว
Multiple regression is a statistical technique that allows us to examine the relationship between one outcome and multiple predictors. It extends the concept of bivariate linear regression, where we model the relationship between two variables, to include more predictors. In the context of social science research, multiple regression helps us answer the question: What is the unique effect of on...
A Primer on Causality for the General Linear Model
มุมมอง 245ปีที่แล้ว
While statistical methods can identify associations between variables, they cannot determine the nature of causal relationships between variables. Causality is either assumed on theoretical grounds, or established using the experimental method. This video introduces the building blocks of causal models, including confounders, mediators, and colliders. Using these building blocks, you can make m...
GLM IV: Categorical predictors or ANOVA
มุมมอง 655ปีที่แล้ว
We can use the general linear model to examine mean differences between the categories of a nominal or ordinal predictor with more than two categories by using dummy coding. We code one variable as the reference group (giving it the value 0), and estimate the mean difference between the reference group and the other categories.This regression model is completely equivalent to one-way ANOVA (Ana...
GLM III: Binary Predictors or the Independent Samples t-test
มุมมอง 450ปีที่แล้ว
We can examine group differences in a continuous outcome variable using bivariate regression, or using the independent samples t-test, which is a special case of the general linear model. This lecture describes differences between two groups from both perspectives, and introduces dummy coding and the effect size called Cohen's D for between-group differences.
GLM II: Sums of Squares, Correlation, and Standardized Regression Coefficient
มุมมอง 577ปีที่แล้ว
This more technical lecture covers sums of squares and how they relate to model fit, as well as correlation and the standardized regression coefficient, which are closely related concepts in the context of bivariate linear regression.
GLM-I: Bivariate Linear Regression
มุมมอง 566ปีที่แล้ว
General Linear Model (GLM) is a family of models used to analyze the relationship between an outcome variable and one or more predictors. In this lecture, we will focus on bivariate linear regression, which describes a linear relationship between a continuous outcome variable and a continuous predictor. We discuss testing model coefficients and the assumptions of bivariate linear regression.
Sampling Distribution and Standard Error (Lecture 3)
มุมมอง 579ปีที่แล้ว
Sampling Distribution and Standard Error (Lecture 3)
Complementing theory on adolescent emotion regulation using machine learning
มุมมอง 2732 ปีที่แล้ว
Complementing theory on adolescent emotion regulation using machine learning
Utalent "Toeval of Ontdekking": Wat is de rol van statistiek in onderzoek?
มุมมอง 1173 ปีที่แล้ว
Utalent "Toeval of Ontdekking": Wat is de rol van statistiek in onderzoek?
Introduction to Structural Equation Modeling
มุมมอง 1.1K3 ปีที่แล้ว
Introduction to Structural Equation Modeling
Workflow for Open Reproducible Code in Science (WORCS), presented at SRCD 2021
มุมมอง 1683 ปีที่แล้ว
Workflow for Open Reproducible Code in Science (WORCS), presented at SRCD 2021
Control Mplus from R using MplusAutomation
มุมมอง 4.3K3 ปีที่แล้ว
Control Mplus from R using MplusAutomation
Verbanden tussen categorische variabelen
มุมมอง 1.2K3 ปีที่แล้ว
Verbanden tussen categorische variabelen
ERRATA: 19:40 - 19:50 I incorrectly report the probability of P(Z > 1) as .025, but it is .16!
Note the ERRATA: 10:10 - 10:50 I talk about the probability of Being Dutch and Having a Tattoo, but I'm calculating the probability of Being Dutch and Not Having a Tattoo (I misread the column labels).
Thanks a lot for this video!!!
This is SOOOOO helpful! Thank you so much
I want to explain why some path coefficients are fixed at 1, thank you
Great
Hi Prof van Lissa! This is an awesome video. I have been looking for a demo and I landed on your website. By any chance, do you have a tutorial or a script on how to do LCA and k-fold cross-validation using MplusAutomation package? Thanks a lot!
Thank you! As LCA is an unsupervised learning mrthod, I don't think cross validation is relevant (what's the error metric when there's no outcome?). Also, you might want to look at cjvanlissa.github.io/tidySEM/articles/lca_exploratory.html
Prof @@CJvanLissa Thanks a lot for answering my query and sharing this material. :) Very helpful.
Thank you so much for creating this tutorial! I can't say enough about how much it helps! I'm also amazed by how clear and fluent your video is. A joy to watch!
Very nice. Thank you
Hi Caspar, came across your tutorial here! I am working on a large scale RI-LTA model and this is very helpful! Thank you! --Terry
This is one of the most well structured lectures I've encountered whilst studying at the UU. Thank you so much for making such high quality educational content
Wauw, mooie uitleg!
Enig puntje: X -> Y is "de regressie van Y op X" (i.p.v. "de regressie van X op Y").
Uiteraard geeft dit wel het effect van X op Y aan.
Very nice tutorial! Any thoughts on how to generate graphs for finite mixture models?
That depends.. there's functionality to plot the distributions of the data in such models in both tidySEM and MplusAutomation (tidySEM is more up to date), or do you mean drawing a graph of the relationships between the variables?
Nu de echte vraag; heeft Caspar de broodjes ook opgegeten?
Was best lekker!
ik wilde net typen: nu de (h)jamvraag...
Do you have courses on SEM or Exploratory data analysis on online learning platforms like Coursera, Udemy, EdX? Your lectures which are taken outdoor are so unique and innovative. Thank you.
Thank you very much, Caspar. I have issues when I run mplusModeler. Error in path.expand(path) : invalid 'path' argument Could you have a comment?
Oh, I see your videos shifted into a positive direction from week 1 until here and they eventually turned out really nice. Appologies for my previous critisism, I did not know previously that these videos are part of the course as well!!!
Great video and very helpful!! I followed through but when running "mplusModeler", the results have just NULLs. Could you offer some possible causes? I do have some missing values in the R data as NAs. Does this matter? I am using Mplus 8.3. Thanks much!
Never mind. It turned out that my model is not converging and I increased the iteration# (10e4) and lessened convergence criterion (0.5e-3) to get the estimates. Please share thoughts if you think it is not appropriate (given a sample size ~3000 and 15 variables forming ~ 5 factors). Thanks again for the awesome video!
Can you maybe also use your mouse to explain the graphs and models from the lecture notes? It would help a lot to follow you while explaining what we need to look at because I got lost at some points.
@barocsai I did the best I could at the time of recording these video's; unfortunately it's not possible to go back and change them after publishing them here
@@CJvanLissa oh I mean in the next videos for now this is good but it always can be better. I hope I am not annoying. :)
Nice video! (I have one small feedback: if you do not use the extra video in the corner when you show the lecture notes then I can focus more on your voice and the lecture notes because in those specific situations I need to focus on: 1, your voice 2, the smaller video 3, the lecture notes. Moreover, it also saves time for you because then you do not need to spend time on cutting or finding the best angle to shot and include within the video. :) )
Thanks a lot Caspar for such a great video. I should be able to work so much more efficiently with this. As for R, it is nice to have someone show which functions to use in which package. :)
Great tutorial! Super clear and helpful:) I had one problem- when running the graph_sem/prepare_graph functions I got an error (for sure my bad :)). Any idea how to fix it? Error in prepare_graph.default(res, layout = lay, angle = 179, fix_coord = TRUE) : Argument 'edges' must have columns 'from' and 'to'.
Yes, there is a known issue with this function. I'm pushing a fix to CRAN this week. Meanwhile, try: remotes::install_github("cjvanlissa/tidySEM") and I think it wil work.
Excellent! Well done
Thank you so much! This is just unblievable, to be sending a question to the www, and be answered by such a detailed and informative video.
Can we do use this package for Mulitlevel SEM?
Yes, the MplusAutomation package just communicates with Mplus so you can use all of its functionality. TidySEM can tabulate and plot multilevel models too.
@@CJvanLissa Thank you very much Casper. How about the Indirect Effect, also workable? Also, can the package read an already existed Mplus output (that generated by Mplus) and then I present the results by using it as R variables? Thank you very much in advance.
@@JZhu-sv8fx yes, use readModels(filename)
@@CJvanLissa Thank you very much Caspar.
Great video! Very useful to use with the ILSA's data.
why did you stop making videos in English?
Thank you for asking; I will continue to make videos in both Dutch and English, depending on the language of the courses I'm teaching.
Making statistiek fun one college at a time
I would be interested in a future video which covers how manuscript revisions are handled in WORCS. i.e. how you track changes (for text and for code), and what the version we re-submit would look like (e.g., is it colour-coded text for changes? is there a separate document for code changes?).
Yes, this is a very good idea. The git system allows you to compare versions easily, including manuscript vs revision or two authors' contributions. When I next have a manuscript under revision, I will document it.
Really amazing! I was looking for a tutorial to help me to connect r studio and github to upload projects and this video is really really wonderful.
Fantastic!
Greetings from EARA 2020 :) Nice package. Seems it has a lot of very useful features and I like it. However, I wonder about the pseudonymisation that is used to change the data. Because the changed data is still similar to the real data. It might mean that if someone will figure out the algorithm for pseudonymisation that was used in your package (the code for which is also available online), they might trace back the real data. Therefore, I’m a bit concerned that ethical committee might be not so happy with publishing such pseudonymisation data online. They can be very tough on this things, especially after GDPR was introduced. Did you had any discussion with the ethical committee about this? Were they convinced it is ok to publish such pseudonymisation data?
Thank you! I think you're referring to the synthetic() function. Synthetic data are generated based on a (random forests) model of the original data. It is impossible to reconstruct the original data from a model, just the same as you cannot reconstruct individual participant data from a regression coefficient - but you can use that regression coefficient to simulate new data. Thus, this is not a real concern.
@@CJvanLissa Thank you for your reply. Is this specified somewhere in your article or a package description on Git? Because, I think it might be nice to show a ref to the ethical committee, if they would have questions like that. ps in my experience they can have a lot of very different questions, so it is nice to be prepared
@@1VideoFromMe yes, as indicated in the WORCS preprint and help files, the algorithm is developed by Nowok and colleagues, doi:10.18637/jss.v074.i11
Awesome research to deliver practical tools