Polynomial regression using SPSS (July 2020)
HTML-код
- Опубликовано: 11 сен 2024
- This video provides a walk-through of options for performing polynomial regression using SPSS. I discuss ways of assessing whether there is curvalinearity between independent and dependent variables and ways of modeling any curvilinearity between variables in a regression model. To obtain a copy of the Powerpoint referenced in the video, click here: drive.google.c...
A copy of the SPSS data file can be obtained here: drive.google.c...
This video is such a comprehensive step by step walktrought. thanks a lot
Dear Mike, Thank you for the video. I appreciate your work. You are a point of reference for learning statistics.
Hi Davide, thank you for your kind words! Best wishes!
This is awesome, making regression down to earth
Thank you, thank you and thank you.
You are very welcome!
That was a good piece indeed. Quick question: why did you use transformed loweff variable? Did the parent variable violate linearity?
Thanks.
Hello Mike, first of all, thank you for making these videos. They are really helpful. I have 2 questions. To start with, my linear regression was insignificant, and for the quadratic one, the second degree variable was insignificant, and the others significant. I assume that since the second degree one is the most important, I should not accept the quadratic regression?
Concerning the part where you checked if there was a cubic equation, after the indication of "tolerance limit for entering variables is reached", I had the same for my cubic regression. I computed a new variable, the mean center one, as you did. Then checked for the cubic regression again. In the coefficient table, all the coefficients were insignificant, except the third degree one. What does that mean? Does a cubic equation exist (since the third degree one is significant)? Or should I not accept it, since all the other coefficients are insignificant?
Thank you for your time.
Dear Mike,
Is multicollinearity a problem for non linear polynomial regression.If I do multiple non linear quadratic regression should I consider multicollinearity
This is very helpful!! Thank you!
Would you please consider do a tutorial on Polynomial Regression and Response Surface Method to test for Congruence/ Fit effects?
Thanks!!
I am also working on congruence/fit model. I think the polynomial regression method followed in the video remains the same. However, to see congruence/incongruence effects on the DV one would need to use the equation we get from the polynomial regression above and plot it or develop an RSM. I used Jeff Edwards excel sheets in the past that are freely available on his website to plot the RSM. The problem gets complex where there are moderators and mediators involved in polynomial regression. I really couldn't find any help on that. I tried using PROCESS model but was not successful either.
Can you model local regression or LOESS? I only see it in the scatter plot. R-squared does not show up for LOESS either?
Hi. Is there any video for moderating effect in polynomial regression ?
Thank you,sir
Very helpful! Thank you so much!
can you use 2 independent variables and 1 independent variable in polynomial regression? just like in linear regression
Thanks a lot
Hi sir, where we can find the RMSE value for this?
Do you have video on interpretation of such relationships
Dear Mike,
Thank you for your educational content, it has been invaluable to many of us! :)
I know this is not directly relevant to this video but I was wondering if you are planning on doing any more tutorials on Canonical Correlation Analysis. I am using the method for my masters thesis and have found that is it surprisingly underused with not a lot of information on troubleshooting online.
Many Thanks,
Archontia
Hi Archontia, I don't have any recent videos on the topic but I do have an older one from 2018 (ruclips.net/video/O_nHyOXPh0I/видео.html). Of course, it probably sounds like I'm in a basement and is not quite as refined as the ones I've put out in the last year or so. However, you might still find it handy. It involves using syntax. I don't know if the more recent versions of SPSS have the option as drop-down menu, but I think you can get an add-on using the extensions hub. If you install correctly you'd find it under the correlate menu. The output is the same as if you use syntax, but it does look nicer. I hope this helps! Good luck with your research!
@@mikecrowson2462 Hello! I am following the video you have shared to conduct my analysis and it has been incredibly helpful so far. Many Thanks!
I am doing my research and encounter this in SPSS. "Cubic model could not be fitted due to near collinearity among models. SO meaning I will chose form Linear or quadratic models ?
You probably need to mean center your predictor first before using curve estimates. I actually have recently made a new video on polynomial regression with a single predictor at the link below
ruclips.net/video/IKHHtwFS_Uw/видео.html
@@mikecrowson2462 My Professor gave the data and I need to perform Polynomial Regression in SPSS and I need to choose the correct model to use whether Linear, Quadratic and Cubic. When I performed the test I encountered that problem. I have two choices left either to use Linear or Quadratic but not sure about this.
I will watch the video. Thank you for the reply. I really appreciate it.
hi sir, how about quartic for 5 treatment?
how to get polynomial ordo 4?
Unfortunately, SPSS does not have a default option for obtaining that type of polynomial regression (or at least I don't know of one). You might be able to run this through the non-linear regression module, but it is likely to be unfamiliar and appears to rely on an interative estimation approach (see e.g., www.ibm.com/support/knowledgecenter/SSLVMB_24.0.0/spss/regression/idh_nlre.html).
HOWEVER, you can easily perform it by thinking in terms of the following polynomial function: f(x) = b1x^4 + b2x^3 + b3x^2 + b4x + b0, where b1-4 are the regression slopes multiplied by x raised to a power & b0 is the intercept. So, to perform the regression, you would need your variable X, then you will need to compute another variable that is the square the x (i.e., x^2), then you will compute another variable that is x-cubed (i.e., x^3), and then finally compute a variable that is x raised to power of 4 (i.e., x^4). Enter all of these variables into your linear regression and you will obtain your model. I would suggest entering the variables in a hierarchical fashion and test for R-square with each step to see which added terms account for significant increments in variance. Finally, since these variables are likely to be highly correlated, you might consider mean-centering (X-mean) your X variable first before creating the higher order terms. Although it's not required, if the program detects very high collinearity, it might exclude highly correlated terms. The mean-centering could be useful work-around to prevent that from happening. Hope this helps!
please do one video on GLM in SPSS