Multiple regression using dummy coding of multi-categorical predictors in SPSS (August 2021)
HTML-код
- Опубликовано: 24 авг 2021
- This video provides a walkthrough of dummy coding of multicategorical predictors in linear regression. I begin with a simple example using gender identification as a predictor of life satisfaction to lay the groundwork for the later demonstrations where I use the multicategorical variable 'marital status' as a predictor of life satisfaction. In this video I also compare results using dummy coding with results obtained by way of independent samples t-tests, one-way ANOVA, and one-way ANCOVA.
Download a copy of the SPSS dataset here: drive.google.com/file/d/1_qnP...
Download a copy of the Powerpoint referenced in the video here: drive.google.com/file/d/1GAX6...
For more videos on regression using SPSS, please go to: sites.google.com/d/1ZuEPh-Z_f...
I have no words to thank you Mike. Just to say you are a star. I have watched many of your videos and I must admit they are more than educative. As a PhD candidate, you made my academic journey easier, at times making me feel statistically inadequate. But, isn't feeling this the best way to learn, isn't it? Thank you Mike
This is awesome! Starting my dissertation and this just simplified everything! Thank you!
Finally found a video that is clear and helpful! Thank you!
Thank you SO much for this video! Your explanation was so clear! Just saved my master dissertation.
Quite elaborate. Just debunked all the confusion I had with MLR with multi-categorical dummies. 👏👏
This is a fantastic tutorial! I struggle with understanding statistics at times, yet you made this very simple to understand. You are a skilled teacher indeed. Thank you so much for creating and sharing this tutorial. It has helped me with my research assignment.
This was so useful for my masters project. Thank you!
Thank you! This has saved me on my thesis analyses!
Thank you SO much for this!! This is hugely helpful.
This was so much helpful. Much appreciate you.
Thank you so much for your excellent explanation! it's really helpful!
Thank you for your excellent explanation...
Thank you. This tutorial is soooo helpful!💗
You're so welcome, Ning!
This is awesome! Thank you very much
Very very nice presentation tanks for all
Thank You so much.
Brilliant explanation 👏
I am glad you liked it! Cheers!
Mike, the indicator var D4, if used (encoded) will generate redundancy and singlarity issues in OLR,....so why one hot encoding is generally used in ML without any problem? what's your opinion on this?
thanks!
Thanks for the video very inspiring and have learned a lot.
Plesse, what is yout dependent variable is dichotomous? And will have to run it
with thr dummy variabled in linear regression?
Thank you so much❤
You are very welcome, Menu!
Cheers!
thank you so much
You're welcome!
OMG thank you so much! I have a question, though... how to dummy code for interaction terms? Specifically, I wonder how to create a dummy code for the interaction between income and maritalstatus? Thank you!
very nice video, Dear Mike Crowson, can you tell me how to interpret one categorical IV marital status under second categorical IV gender id with respect to life satisfaction and income if they are not normally distributed?
This video was so helpful!! Does anyone know how i’d report results from this type of analysis in APA format?
Very helpful video! I am wondering if we have to make any adjustments if we use two or more multi-categorical predictors in the regression model that we have already dummy coded?
Hi there. No, I don't believe there would be any adjustments you'd need to make to include multiple categorical predictors in the regression. Cheers!
@@mikecrowson2462 if this is the case, then can you pls clarify how to write interpretation and also what all values should we report in the writeup
Thanks for your explanations. I would like to know what to do with the level of measurement. My categorical data have been assigned (string value) nominal data and when I create a dummy variable, should I change the level of measurement to scale?
My general recommendation would be to convert your variable to nominal or ordinal and then use the dummy variable coding option (i.e., Transform-->Create Dummy variables) discussed in the video.
Also, I don't know if you'd be interested, but I will also be covering this topic in a larger seminar I will be putting on in December on regression using SPSS and Process: instats.com/seminar/regression-and-moderation-analysis-with-9776
Cheers!
But if you put on the model there will be colinearity variables .. because individual could not be single and married at the same time , right?
I'm not really clear on what you are asking. Based on your question, it sounds like you may be confusing the dummy variables with the categories themselves. The full set of dummy variables are used to establish which group a given case is a member of. They are not the actual categories such as married, divorced, single, etc. When they are entered as a set into the regression model, the regression slopes for the dummy variables represent comparisons on Y between individuals that happen t fall into a given group and a reference group. Collinearity is a non-issue here. [The only way for collinearity to become a problem would be if you created four (instead of 3) dummy variables here and entered all 4 dummies as predictors in the model.]
can we consider all the three variables i..e, gender, marital status and income at same time for regression.
Yeah, I've got the same question. The video is really helpful and gives clear explanations, but I still don't know the answer to this because it's related to the interpretation and report. Can we analyse gender (dichotomy), marital status (dummy), and income (numeric) all at once in an analysis? Please let me know if you got the answer.
Very informative video sir.
If correlation coefficient is positive but regression coefficient (multiple regression analysis)of the variable comes negative what inference can be drawn from such analysis? Kindly tell.
Hi there. What you are describing sounds like a suppressor effect. The zero order correlation between x1 and y is positive (and the regression coefficient would be positive in simple regression). If you add x2 to the equation, that previous slope becomes negative. This is a type of suppression effect that is often observed in regression analysis. I just ran across this blog posting by Allison that does a nice job of describing this: statisticalhorizons.com/when-do-suppressor-effects-occur/
@@mikecrowson2462Thanks for replying. So, is it fine to have suppressor effect in multiple regression analysis. If not ,how can we present our results in research papers.
There's no problem with the suppressor effect occurring in your analysis apart from making interpretation of the 'counterintuitive' sign for your variable. Sometimes, you can come up with a theoretical claim as to why it occurred. Other times you may be left with simply indicating to the reader about the presence of suppression with not much more.
@@mikecrowson2462 Yes sir, I was thinking the same .As i am not a statistician it becomes complex for us to explain such analysis.
can multiple regression be applied with dummy variables ( my data covers a period of 13 years). can I apply multiple regression with 12 dummy variables (year) along with other variables like age of company, issue price, issue size etc? how many independent variables can be selected for applying multivariate stepwise regression? Sir can you please share you email Id If possible
Thanks so much. But please why did you choose linear instead of multiple regression among the options when we have various predictors regressed on a single outcome? Thanks
thank you so much
You're welcome!