IRT Models (Rasch, 2PL, & 3PL) in R with ltm package

Подписаться 953

Просмотров 9 тыс.

50% 1

☕If you would like to support, consider buying me a coffee ☕: buymeacoffee.com/statsguide8
For one-on-one tutoring/consultation services: guide-tree-statistics-consult...
I offer one-on-one tutoring/consultation services for many topics related statistics/machine learning. You can also email me statsguidetree@gmail.com
For rcode and dataset: gist.github.com/musa5237
Tutorial of how to fit three unidimensional binary response Item Response Theory (IRT) models (Rasch, 2PL, and 3PL models) using the ltm package in R. The dataset used were responses to the 25-Item online IQ test modeled after Raven Progressive Matrices (RPM).

Опубликовано:

13 дек 2020

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 25

@statsguidetree 3 года назад

Here is the rcode: #where to download IQ1 dataset # openpsychometrics.org/_rawdata/ #load dataset download IQ1

@ikasurtiani4293 2 года назад

Thank you, this is very useful for me. But, when I try to do 3PL model analysis I got this message "Hessian matrix at convergence is not positive definite; unstable solution". What does it means and what should I do? Thank you in advance.

@statsguidetree 2 года назад

@@ikasurtiani4293 Hello Ikra, this could be do to the model failing to converge appropriately. The model needs to converge in order to produce parameter estimates. In the function ltm() or tpm() include the argument start.val = "random". This may help with convergence because it ensure starting values used in the estimation process are random. You may also want to set.seed() to some value before running the your function so the same parameters can be produced.

@maciejbienkowski7488 Год назад

Awesome job! Thank you for doing it, it's very helpful.

@ilusayatsani8091 3 года назад

Thank you, this has been massively useful to me! You're awesome

@statsguidetree 3 года назад

I'm happy you found it useful.

@christianodinga2289 Год назад

Thank you so much for this video! I hope I can get your consultation as I work on my analysis

@user-ek5rc5kl7o 2 года назад

Thank you for this wonderful tutorial. I wonder if I have adopted the Rasch model, can I still use the 2PL model function to check unidimensionality? Thanks!

@hanyiskandar5567 3 года назад

Great job! thanks alot!

@statsguidetree 3 года назад

I'm glad you found it helpful.

@chaimatibajjate3751 Год назад

Thank you for the video. Please, I have a question about "factor.scores()", it doesn't work for me because I receive an error of a missing argument "f". Do you have any solutions to solve this problem ?

@user-kx5sd8tz2r Год назад

Amazing videos! Just a question: I have a questionnaire with a 5likert scale. Shall I use the Rasch or GRM model?

@statsguidetree Год назад

For likert scale data you generally want to use a GRM.

@mariyamirza 3 года назад

I would appreciate it if you please tell me the r codes to estimate the students' ability by fixing the item parameters (difficulty ad discrimiation) in 2PL model.

@statsguidetree 3 года назад

Are you referring to attaining Composite scores for the ability estimates? If that is the case you can generate composite scores for each examinee using the following rcode (where IRT2pl=name of the 2pl model and IQbin=name of the dataset). # ability estimates with Component est

@Rocky_E 3 года назад

Really helpful video, thank you. When i run my data through all 3 models the probability of getting the item correct doesn't range between 0-1. All my numbers are greater than 1, some as high as 9. do you know what this means? how can i fix this.

@statsguidetree 3 года назад

That is interesting. It could be a couple of things. Would you be able to send me a snap shot of the output results to my email statsguidetree@gmail.com ?

@almalen2784 2 года назад

do you have any tutorial on how to test irt assumptions using R?

@statsguidetree 2 года назад

Currently, no. But, I can do something in the very near future. As for now, there is the unidimTest() function in the ltm package that you can use to check the unidimensionality assumption of an IRT model you generated e.g., unidimTest(mod1). I will be sure to notify you if I do post another video in the near future going over assumptions (e.g., Unidimensionality, local independence, monotonicity, and item variance, etc.)

@khashayarasgari3226 2 года назад

Could you please let us know how to run 4PL model ? Thanks

@statsguidetree 2 года назад

I do not believe there is a way to do 4PL models in the ltm package -- I could be wrong. If I post another video on IRT, I will try to do a review and see if there are other packages available that conduct the 4PL.

@Dupamine 3 года назад

Can you also make tutorial on multidimensional irt?

@statsguidetree 3 года назад

That is a good idea. I can try and post a video of multidimensional IRT in the very near future.

@goodboy-je2kz 10 месяцев назад

What about SEM ? Can we do it in this package ????

@statsguidetree 9 месяцев назад

I am not too familiar with that package. There may not be a specific function for irt models but it may have a more round about way of estimating irt parameters.