Standard multiple linear regression on SPSS with assumption checks and write up (APA style)
HTML-код
- Опубликовано: 6 июл 2024
- Learn how to complete a standard multiple linear regression on SPSS with assumption checks and how to report the results in APA style.
0:00 What is multiple linear regression analysis?
0:46 Setting up the file
2:01 Running the analysis
2:59 Checking assumptions
7:25 Interpreting the regression results
9:07 Reporting the assumptions
12:14 Reporting the regression
Regression equation:
The B values can be used to predict the value of the dependent (outcome) variable. In this example, the B values were 0.616 (distance from campus) and 0.375 (distance from nearest train station). As well as the B values, we need to know the constant value, which is 0.070 in this case (see Coefficients table). If someone lived 5 miles from campus and 5 miles from their nearest train station, the model would predict that this person would be late to lectures about five times: (5 x 0.616) + (5 x 0.375) + 0.070 = 5.025.
![mgk - I Think I’m OKAY (with YUNGBLUD) (Sad Version) [Official Music Video]](http://i.ytimg.com/vi/7AC1uqQLIg4/mqdefault.jpg)
Thank you so much! So many things that I have been studying for ages, suddenly make so much sense. I am very appreciative.
Thanks Philippa, glad it helped!
Thankyou very much for this video David, was a huge help in my Stats psychology assignment! Also really like how you incorporated your written report at the end, big help. A lot of other videos i've seen only describe the SPSS procedures, having the written side explained helps heaps in assignment settings. Cheers
Thanks Christian, glad it was useful!
Dr. Robinson, your step by step handling of assumptions was extremely helpful and concise. Thank you for such a clear cut video on MRs.
Thanks, glad it helped!
Thank you Dr. David Robinson. So great to see such a detailed video. The way you smoothly teach and narrate your analysis is amazing.
Thanks Naurin!
Thank you so much for this clear tutorial. I am just writing up my assumption checks for a MLR for an assessment (BSc Psychology), and this helped enormously.
Thanks Jane, glad it helped!
Thanks
Your video has been so helpful. Thank you!
Thanks Konstantinos, glad it helped!
Thank you so much! You explained everything so clearly and I actually understand it now! :)
Thanks Emma, glad it helped!
Awesome, thank you. I cannot express how much this video, and your breakdown of each of the steps, is appreciated! #Subscribed
Thanks Michael!
This was extremely helpful. Thank you!!!
Thanks Katie, glad it helped!
Great video, very well put together and explained. Huge ++
Thanks Shane, glad it helped!
Thank you you saved my life :)
It is super helpful. Thank you so much!
Thanks, glad it helped!
This video saved my life. Thank you so much
Thanks Simon, glad it helped!
@@DavidRobinsonPhD One thing I am curious about with Cook's distance is whether one should not also look at the individual values. I ran the regression exactly like you showed and overall Cook's distance was in acceptable range. However, when I included case wise it flagged one case as problematic.
btw i'm undergrad cambridge and you help more than my supervisors
Thanks Mia, glad to help!
Thank you so much for this insightful video! What about how to write the hypothesis final result itself after reporting the Multiple regression results. Like should i do it after or at the same time as I report say one of the predictors. If so, how in APA format?
Thank you unlimitedly for your informative explanation!
Would you share with me a word copy of the file used in reporting those results ?
HI, thank you! a very clear presentation! I have one question: do I need to check assumptions with two IV's only (if I have two IV's only) before introducing an interaction term (before running process macro) or do I have to check assumptions with an interaction term as a third IV? Thank you!
Hi David! Love your videos. Can you also do a video on SIMPLE LINEAR REGRESSION with assumptions checks and APA write-up?
Thanks Risa! I'm planning to make some new videos within the next couple of months, so will try to include that one!
Tahnk you so much. That was very helpful. He deserves "subscribe" (:
Thanks Abdulaziz, glad it helped!
Hi Dr. Robinson, great video. Just wondering what do you do if your r value is closer to zero? Does this mean that regression isn't a suitable test? Thanks!
Hi Mary, thanks for your question. It's not usually considered problematic if the r value representing the relationship between the predictor values is close to 0, but the assumption of multicollinearity may have been violated if the value is above .7 (or further from 0 than -.7).
Great video! It would be great if you could show how the tables are made in APA format
Thanks Anushka! Yes, I should have shown how to do that. There's a good example towards the bottom of this page: dyudkin.medium.com/custom-r-regression-functions-you-might-find-useful-8f58d610f41
@@DavidRobinsonPhD Thank you! That was very helpful.
Great learning Vids. Can you do one on MANOVA?
Thanks Harry! Actually planning to record a MANOVA one this weekend!
Hi, David Thanks for this great video. I am performing a hierarchical multiple regression. My sample size is 1600. Is it still important that all assumptions be met?
Hi Sheena,
Thanks for your question.
The same assumptions would apply. If there are any minor violations of these, you could consider running the analysis anyway, but perhaps acknowledge the violations in your report.
Hope that helps!
David.
Thank you David. Just one question if i have a P value which is highly insignificant such as .542, would I report this value or would I simply state that p>0.5 ?
Thanks :)
Hi Ertan,
Thanks for your question. APA guidelines say to report exact p values (except when they're very small, in which case p < .001 can be used), so I would report that as p = .542.
All the best,
David.
@@DavidRobinsonPhD Thanks a lot David!
Hi, what if the assumption for multicollinearity has not been met? should we still perform multiple linear regression?
Hi David,
I have two independent variables in my study and tried using multiple regression, but all the assumptions are not met. Can I still show it in my research and conclude with the results?? OR I have to make sure that all assumptions are met??
Hi Pratik,
Thanks for you question.
It's possible that you could still report the results. If you do, you could try to find literature that supports the argument that the results are still valid even if certain assumptions are violated. For example, in the case of ANOVAs, normality is an assumption, but some researchers (e.g., Julie Pallant) argue that mild violations of this assumption aren't problematic. You could also highlight in the discussion that assumptions were violated and inform the reader that the results, therefore, should treated with caution.
Whether it's OK to report the results despite the violations depends on many factors (e.g., how many violations there are, how severe they are, where the work will be shared), so if you have a supervisor you can get advice from on this, it would probably best to speak to them.
All the best,
David.
Hello, i was just wondering if it was possible for the model to significantly predict the outcome variable but when looking on the coefficients table, only 1 of 3 variables be significant? Thank you!
Hi Ella, thanks for your question. Yes, it’s quite common for the overall model to be significant and for only one of several independent predictors to be significant. This just suggests that that significant variable is what is contributing most to the predictive power of the model.
@@DavidRobinsonPhD Thank you so much! You’re a life saver!
Hi, why did you only take 1 value from the Pearson correlation for testing multicollinearity? I'm conducting my own multiple regression for my dissertation. I have 1 extra predictor variable than you and am struggling on which value I report! Thanks :)
OMG, I had the exact same question. This video has been so helpful but I'm unsure how to report Pearson Correlation for 4 predictors. Please help!!
Great video! Can you do one on Exploratory factor analysis?
Thanks Grace! I've got a list of videos I'd like to make over the next few months, so I'll add it to the list!
Hi David thanks for this. I still have a few questions and wondered if you have a contact email? No worries if not but finding your explanations helpful.
@@gracegibson5443 Feel free to ask any questions you have here, as this might be useful for other viewers. However, there is an email address in the channel description.
@@DavidRobinsonPhD would this video be an example of a quasi-experimental design, or is it a correlational observational study. Would checking the Pearson's co-efficient be the best way to express the strength of relationship between the variables and to indicate the significance level?
And lastly when you write up your results. What is the difference between reporting the assumptions/ the multiple linear regression analysis (to estimate each variable’s relative contributions to the outcome ) and talking through what you could conclude from the analysis. Like, can you add anything other than just saying distance from campus and train station impacts how often they are late to lectures, without repeating the regression report?
@@gracegibson5443 Hi Grace,
Thanks for your question.
It’s an example of a correlational observational study.
Yes, Pearson coefficients would probably be the best way of reporting the strength of the relationships between variables (with values further from 0 indicating stronger relationships).
When writing up, the results section is usually just used to simply state the results of the analysis. However, in the discussion section you would interpret the results with reference to similar studies and relevant theories.
Best wishes,
David.
Hi David, I was wondering when you would report your screening/cleaning process & descriptive statistics in this result section?
Hi Hannah, screening would usually go first, followed by descriptives, then inferential statistics (e.g., regression).
@@DavidRobinsonPhD That’s very helpful David, thank you! Have you done any more videos in relation to these 2 earlier processes for multiple regression? 😊
@@hrosem Not yet, unfortunately, but will aim to do that sometime.
@@DavidRobinsonPhD Hi David, sorry last question I promise! For the multiple regression, when reporting the descriptives (M & SD), what do they actually tell you? Like how do you comment on the means for questionnaires total scores. I know you can say one mean of a questionnaire total score is higher than the mean of the other questionnaires total score. But what are these means actually telling you?
I understand the SD, as that tells you how much participants varied in their answers of the questionnaires.
But I am just struggling to understand how reporting the mean is helpful in understanding what's going on when dealing with 3 questionnaire total scores mean values!
Hope this means sense! THANKS! :)
@@hrosem Results sections often say very little about the means and standard deviations, other than they were calculated. However, if the means can be used to categorise participants, you might be able to briefly comment on which category the mean is representative of. E.g., mean scores on a measure of depression might reflect mild or severe forms. Reports may say more about descriptive statistics in the discussions, such as by comparing means for the current sample to means obtained for other samples in past studies.
How would you proceed if you have an r value of exactly 7 but acceptable VIF and tolerance values? It seems that the assumption has been met despite the high r value given the other data.
You could consider removing one of the predictors or combining the two variables into one. However, this probably isn't necessary as .7 is quite a strict threshold (I've seen .9 suggested elsewhere).
@@DavidRobinsonPhD Thanks a bunch, very, very good videos by the way.
you should be a psychology teacher
you missed the linearity assumption of multiple linear regression model.