SPSS - Hierarchical Multiple Linear Regression

[@audtracol]

Thank you so much! You have a wonderful way of explaining everything! Really appreciate it.

[@lacrimosa2994]

Thank you so much, this video is a lifesaver! I'm wondering about reporting R² alongside pr²... they are slightly different values, how can it be that the % change can be both 50 and 55? I know you said that they don't add up... But when reporting, if i want to say that the addition of a variable was responsible for change if variance in the overall model (R²= .50) but the pr² vaule for this variable in said model is pr²= 55... which one should be reported as percentage change?

[@StatisticsofDOOM]

I usually report R2 values with the F statistic because it's part of the overall model. Then I report pr2 values with the coefficients they correspond to - this hopefully helps keep it clear what each effect size matches.

[@lacrimosa2994]

@@StatisticsofDOOM yes that makes sense, thank you so much! Also thanks again for the video, you are saving many lives 😁

[@jarnoduwe7071]

Why did you use the beta, t and p for sex and age from model 1 and not from model 2?

[@StatisticsofDOOM]

I generally talk about the variables in the step they are entered, rather than in each step.

[@thudang5603]

This video is amazing. Thank you.

[@StatisticsofDOOM]

Thanks for the kind words!

[@miao9732]

Dr. Erin, thank you for this video, I would like to ask whether the dependent variable should be normal in regression analysis, if the DV is not normal, should I conduct linear transformation to improve normality before doing regression? or if the sample size > 30, the normality is not a big issue? hope to hear from you, thanks

[@StatisticsofDOOM]

The sampling distribution should be normal, so yes at least a bigger sample or the DV should look normal (you can also test the normality of the residuals).

[@miao9732]

@@StatisticsofDOOM Dr. Erin, thank you for your reply, I will follow your video to test the normality of residuals.

[@russyallop3957]

Great video, thanks

[@consciousness88]

Thank you so much! regarding Model 2, would you interpret the value same way as in model 1? as extro goes up care for cars goes up? thanks!!

[@BreakTheRules2011]

This is a fantastic video, but I am a little confused. I have tried searching the web for the answer and I really can't find it. For my post-grad coursework, I have a dataset with 1566 respondents on which I am conducting a hierarchical regression analysis. Some of my control and predictor variables are categorical (sex [f/m coded 0-1], age [11-16 coded 0-6], distance to grandparent [coded 0-4], living with grandparent [0-1], free school meals [y/n coded 0-1]) and my other 3 predictors (concerning grandparent influence) are continuous/scale. As you see, I have dummy coded the categorical predictors, but how do you include and present them in a table of correlations? :'( - I can only find examples of this with all continuous/scale data, as in your video. The correlation tables don't make much sense when including each dummy variable on its own merit. Would recreating the original predictor variables, removing the missing values, etc. and making them all scale/continuous be acceptable if I then used those singular variables in a correlation matrix? Finally, for the values that I am not entering into the hierarchical regression (all dummy variables I have coded as '0') should these also be included in the correlation? My head is about to pop! Thank you so so much to anyone who can reply and help me out with this

[@StatisticsofDOOM]

I don't know what I would present correlations with dummy coded variables....usually you might present means for each group? Is that what you are asking?

[@evali5895]

Hi, thanks for all your videos, they are great. I am wondering if can we do hierarchical regression with PROCESS macro?

[@StatisticsofDOOM]

No, you cannot.

[@Chelseabea]

So if Model 1 (age and sex) is not significant but Model 2 is, would you still report the change in F value and R square change for model 2? Or because R squared was not significant for model 1 is there no change in it? Also, if the control variables are not significant in Model 1 but are when in model two - how woud you report that? Is it due to a shared variance between say age and a predictor?

[@Chelseabea]

+Erin Buchanan Ah right, thank you! That really helps :D

[@anthonybrown2200]

How would you report out if the beta for Sex was -0.68 instead of 0.68 in Model 1?

[@anthonybrown2200]

+Erin Buchanan Thank you!!!

[@whitenia87]

Interesting! Thank you so much Dr Erin. You just helped me. However, can I request if you could do the mediating effect for 1 mediator with 2 or 3 predictors using spss or could it be possible to explain the mediator effect manually using this SPSS hierarchical multiple regression on the coefficient output of Model 2? Appreciate your feedback. Thank you, Dr Erin. You are angel.

[@StatisticsofDOOM]

I have multiple videos on mediation - are you saying without something like the process plug in?

[@edgaallan8962]

thanks for that clear lecture :-)