Variables and Factor Analysis
HTML-код
- Опубликовано: 15 июл 2024
- This is a video lecture for the virtual residency for the DM class at CWRU. This lecture covers the various types of variables, specification, and an intro to exploratory factor analysis.
Here's a fun pet project I've been working on: udreamed.com/. It is a dream analytics app. Here is the RUclips channel where we post a new video almost three times per week: ruclips.net/channel/UCiujxblFduQz8V4xHjMzyzQ
Also available on iOS: apps.apple.com/us/app/udreamed/id1054428074
And Android: play.google.com/store/apps/details?id=com.unconsciouscognitioninc.unconsciouscognition&hl=en
Check it out! Thanks!
Probably one of the best educational videos I've found on the internet, ever. Unlike many others this wasn't a waste of time. Thank you!
Great examples of reflective vs. formative variables! Great videos, Gaskination! We are very grateful for all of them!
Wow. I've been struggling to learn this stuff on my own for 8 months, and just came across your videos. My gratitude is boundless. You must be an excellent teacher. (In addition to being a nice guy.)
What a fantastic video! I have recently come across factor analysis and your video was a god send! Brilliant teaching!
I am a student studying business administration on my 3th semester, currently I am very busy with many subjects and courses, thank you very much for giving me the possibility to catch up using your videos! Regards from Denmark
Thanks for the fast reply, you are a life saver. I have only worked with SPSS, but I am going to give PLS a shot :)
the video clip which you used in explaining factor analysis is quite remarkable
Firstly an immense thank you for your vids and the responses to my questions. Alas, another question...
Thanks! Love the positive feedback. I think I'll make some more :)
A fine exposition. Students should watch this and follow ups.
thanks! i'm designing a questionnaire and this helps!
Bee Rain Most of my slides, or variants of them, are available on my wiki: statwiki. kolobkreations. com
Wonderful!
this is very relevant material
Good work!
So great explanation!! tks
and by the way ..enyojable
Thanks!!
@catsfancyful
Glad you like it! You may also find my wiki useful. It can be found at statwiki. kolobkreations. com
Enjoy!
Formative measures do not undergo traditional EFA in SPSS or CFA in AMOS because the items are not required to be correlated within the factor. You can use PLS for formative factor analysis. I have a video about this called: "smart pls factor analysis". Hope this helps.
Dear Dr Gaskin,
Thank you very much for all these awesome tutorials! I was wondering, what is the answer to the question of the assignment at the end of this video? I am working with formative scales for the first time ever and was wondering if this alters the factor analysis itself or only the interpretation of the outcomes. And how it alters the interpretation. I tried to find more video's about formative scales, but you really have a huge collection and I don't know where to search :)
Best,
How is the JOB scenario there related to Data Analytics?
Please make your slides available for downloads.. great help!
are there any rules of thumb re. how much variance a good efa solution should explain.. ( I'm running a series of efas using a range of forced factor solutions, rotation mtds, and have obtained solns where the original 73 items have been reduced to around 30 items -by eliminating low loadings, and cross loadings, etc.) in additon is there any way of determinign how much variation an oblique roation such as promax explains?
@CandidlySubtle Comments cannot contain urls (links), so you'll have to type it in yourself (shouldn't be too difficult...). Sorry. Just go to my Gaskination channel.
Dear Dr.Gaskin,
PLS indeed works really well with formative scales, but now I came across a different problem. I have a model with a moderating variable, namely I am testing the effect of market diversity on the relationship between explorative learning and exploitative learning, and I can't seem to model this in PLS. Is there an easy way to solve this?
Kindest regards, Sonja
Hi Dr Gaskin,
Grateful if you could help for the following:
I am analysing a model-Extended Technology Acceptance Model. For instance, I have a construct named PERCEIVED USEFULNESS, and under that construct, i have multiple questions. I would like to group all those questions under one single variable. I've got like 5 constructs altogether and each of them have 3 questions. I would the need to do a reliability test. I am really bad at spss. Grateful to provide your help.Thanks
Can you post the demonstration links at 13:37 to the information box or whatever it is called?
@PRFDK
You're welcome! Good luck! Try my wiki: statwiki. kolobkreations. com
Thanks for sharing, could you explain for me. The dummy variable included 2 value, but why it's always represented by 1 & 0, but not 1 & 2, or maybe 3 & 4 or 45 & 56? Thanks!
Convention. It is standard practice to use 1 to represent: true, present, on; while zero represents: false, absent, false. This is a very old tradition starting centuries ago, and still used in computers (binary code).
1 & 0 - the binary code. I got it, thanks!
Vielen lieben Dank ! Grüsse aus Zürich
Sorry, I have both nominal and ordinal variable in my questionnaire, i took items from previous stuies and translate from english to french. Which one should i choose between EFA and CFA to check reliability and validity of my questionnaire? or should i use both of them? Thanks for your reply.
EFA is good for exploring the validity of the measures for reflective latent factors. Usually these factors don't include nominal measures. Nominal measures are usually used as grouping variables rather than as part of latent factors. If a variable is not part of a reflective latent factor, then it does not belong in the EFA. The CFA is used for mostly the same purposes, except that the CFA allows you to specify which variables should be grouped together.
Good evening Sir,
I am watching your videos regularly.
Can please upload some videos on Formative scale.
How to built measure construct on formative scale.
And built a Model.
Hi Vikram, I should probably do that sometime. For now, the simple answer is that formative measures should be done at the 2nd order level, with all first order dimensions measured reflectively.
I have videos that show how to do this as a multigroup moderator or as an interaction term. Go to my channel by clicking on my name. Then go to my SmartPLS playlist to find the videos.
what to do when our independent variables are of mixed scale…. For example i have likert scale responses and also discrete( 2 or 3 scale) responses…..
is there any way to perform PCA/EFA on such mixed data?????
do we need to standardise our data?
EFA is usually reserved for reflective factors only. It is uncommon for discrete (2-3) scales to be part of a reflective factors. So, there is rarely a good reason to include such variables in an EFA. The minor exception would be if the short scale is still ordered. For example, if you had education or age group. Education is ordered usually, something like: primary school, secondary school, diploma, some college, undergraduate, graduate, or doctoral. Age is sometimes bucketed into young, middle-aged, and old. These are ordered, so they have the opportunity to covary, which means they might work in an EFA.
got it….tysm for replying back
good
Thank You Sir, for this video lecture. I have a question that can we include categorical and continuous both type of variables in factor analysis ?
Categorical variables do not work well in a factor analysis, because a factor analysis assumes the variables within a factor are correlated. However, categorical variables do not have numerically meaningful variance, so they cannot covary with other variable in a meaningful way.
@@Gaskination thank you Sir. 😃
@James Gaskin Can't we perform a factor analysis using Weighted least square mean and variance adjusted estimation for categorical variables.? Thank you Sir.
@@amanrastogi5184 hmmm... I'm not sure about that one. I don't know much about weighted least square mean and variance adjusted estimation.
@@Gaskination oh, well that's ok. Not a problem. Thank you Sir.
At 9:50 isn't mis specified wrongly labeled as such? I would think that mis-specification happens in 28% plus 1% of the total cases..
Watch my videos on this. I show exactly how to do it. The best thing would probably be to watch my SEM Series playlist. Click on my name, go to my playlists, and find the SEM Series one.
How do you tell a test is unidimensional or multidimensional? Help please!
The EFA will reveal the dimensionality of factors. Here is a video on it: ruclips.net/video/VBsuEBsO3U8/видео.html
Hi How can I download this video lecture which is very important
RUclips used to have a download feature, but doesn't any longer unfortunately. Instead, you can google "youtube downloader HD" and you will find the tool you need in the first search result.
Hi this is very nice. Do you have powerpoint presentation that is included in this video to share? If so, can you please pass it on to me? would be highly grateful
@sajidanaz please email me separately so that I can email it to you. james.eric.gaskin at gmail.com
How to replace missing value?
If you have less than 10% missing (either for row or column), you could replace with the median. Go to Transform, replace missing values. I think I have a video for this. It would be called Data Screening.
So factor analysis assumes you have multiple parallel systems which you use to generate outcomes... Interesting..
I copied this directly from their paper.
Unnecessarily formalized.