Thanks for this video, it's really useful. one question, when looking at the Tolerance and VIF scores for collinearity, should I be looking at the scores of Model 1? Or the best fitting model?
Hi Erin, In terms of effect sizes, should we report Beta or the partial/semi-partial R^2 value? Part of my assignment is making a judgment about which of my variables has a greater 'influence' on the DV, and I am not sure whether I should be interpreting Beta or semi-partial R^2 (or both?) in my analysis. Thankyou!
Hi. Thanks for this explanation. Just a quick question. Why do you check only the normality of residuals for the assumption checking? Does it not necessary to check the normality of the score distribution of all variables?
Not really! But you can - the assumption is the normality of the sampling distribution - if you have a large enough sample (N > 30), this usually ensures that the test statistic is robust to violations of normality.
Hi! Thank you for the perfect explanation! I would like to ask if you have any example when using categorical data as predictors. In JASP we need to code it as dummy variables or there is any other way to indicate that the predictor is nominal? Thank you
I am fairly sure you can just make sure it’s coded as categorical in JASP and it will naturally dummy code it for you (since it’s running R in the background).
Many thanks for your useful video. Could I ask whether we analyze nominal or ordinal data (gender, geography for instance) in multiple linear regression? and how can we do that if yes?
You can! You would need to make sure they are indicated correctly in the variable type option, and the regression will be adjusted for them - check out some information on dummy coding to learn how that works.
Hi Dr. Erin! Thanks for the video! What is suppose to do if p-value at Darbin Watson Test is significant? what's to do next? can you help me? thank you so much!
For the multiple regression, somewhere during 32:13 time, I didn't understand why we wont consider heart rate, could you please clarify it again? Thank you kindly, Sucheta
@@suchetabhuisi318786 Yes! So if you are using alpha < .05 (i.e., p < .05) that specific predictor was not p < .05 but rather p = .187 indicating it was not significant.
@@StatisticsofDOOM Hi, first of all thanks for the lesson! I had a similar doubt. Since the heart rate was not statistically significant, when I report the multiple regression equation, can/should I omit it? Sorry silly question, I'm just starting out in this field. Many thanks from Brazil!
@@StatisticsofDOOM I have 3 IV’s let’s call them A, B, and C. A and C are showing a correlation of -.584 and their tolerance value is .654 and .650 respectively. The tolerance value for B is .951, so my guess is A and C are showing multicollinearity. Hence, I dropped variable C and re-ran the regression with A and B IV’s and found that their tolerance level was same so, I am interested in know when the tolerance levels are same, are variables showing multicollinearity. Thanks!
H0 is the model with no predictors and just the intercept, while H1 is the model with the predictors. You want the model with the predictors H1 to be better than a model with no predictors H0.
Great video! Very clear and easy to follow. I'm feeling very confident as I write up the results section for my thesis.
Thanks for the kind words!
So clearly explained - thank you
Thank you!
Just did my full analysis following your awesome tutorial! Thank you so much!
Glad to help!
Thanks for this video, it's really useful. one question, when looking at the Tolerance and VIF scores for collinearity, should I be looking at the scores of Model 1? Or the best fitting model?
The model you want to use!
Hi Erin,
In terms of effect sizes, should we report Beta or the partial/semi-partial R^2 value?
Part of my assignment is making a judgment about which of my variables has a greater 'influence' on the DV, and I am not sure whether I should be interpreting Beta or semi-partial R^2 (or both?) in my analysis. Thankyou!
Both wouldn’t hurt!
@@StatisticsofDOOM Ok great- thank you!
Hi. Thanks for this explanation. Just a quick question. Why do you check only the normality of residuals for the assumption checking? Does it not necessary to check the normality of the score distribution of all variables?
Not really! But you can - the assumption is the normality of the sampling distribution - if you have a large enough sample (N > 30), this usually ensures that the test statistic is robust to violations of normality.
Thank you so much for the video! What does it mean if my results show correlation but insignificant multiple regression results?
The individual variables are correlated but after controlling for all other independent variables, the individual IVs do not predict Y.
Hi Dr Erin, Does JASP able to perform GMM, 2SLS, 3SLS regression? If so, do you have a guide how to do them?
The newer versions might … but I don’t think it has this capability yet.
Really helpful. Thank you so much 🙏
Thank you!
Hi! Thank you for the perfect explanation! I would like to ask if you have any example when using categorical data as predictors. In JASP we need to code it as dummy variables or there is any other way to indicate that the predictor is nominal? Thank you
I am fairly sure you can just make sure it’s coded as categorical in JASP and it will naturally dummy code it for you (since it’s running R in the background).
Many thanks for your useful video. Could I ask whether we analyze nominal or ordinal data (gender, geography for instance) in multiple linear regression? and how can we do that if yes?
You can! You would need to make sure they are indicated correctly in the variable type option, and the regression will be adjusted for them - check out some information on dummy coding to learn how that works.
Hi Dr. Erin! Thanks for the video! What is suppose to do if p-value at Darbin Watson Test is significant? what's to do next? can you help me? thank you so much!
Not sure I've ever had this problem - check out this website: www.statology.org/durbin-watson-test/
Great video! Thank you very much! May I ask a question! How can I control for a variable when I am conducting a linear regression in Jasp?
You would just add it to your model by putting it in with the other variables, like we did with age/weight here.
How do you do follow up tests for 2x2 factorial multiple linear regression interaction effect using vectors?
I'm not sure I understand what you are asking exactly. I have videos on my channel for moderation if that's what you mean.
For the multiple regression, somewhere during 32:13 time, I didn't understand why we wont consider heart rate, could you please clarify it again? Thank you kindly, Sucheta
I don't think I totally understand the question - are you asking why heart rate isn't useful/significant?
@@StatisticsofDOOM Yes! why didn't we consider it? Why was it not useful?
@@suchetabhuisi318786 Yes! So if you are using alpha < .05 (i.e., p < .05) that specific predictor was not p < .05 but rather p = .187 indicating it was not significant.
@@StatisticsofDOOM Hi, first of all thanks for the lesson! I had a similar doubt. Since the heart rate was not statistically significant, when I report the multiple regression equation, can/should I omit it?
Sorry silly question, I'm just starting out in this field. Many thanks from Brazil!
It was very helpful, thank you very much!
Glad it helped!
Thanks for the video, really enjoyed it
Glad to hear it!
@@StatisticsofDOOM I also emailed you with couple questions. Would much appreciate your feedback. Thanks!
Multicollinearity at 13:14
Explanation at 16:55
All assumptions at 21:50
Hi Erin, this video is really good. May I ask if the completed word doc is available for download?
Everything from the video can be found here: osf.io/t56kg/
Hi Dr. Erin, I wanted to ask what ask does the data show multicollinearity if the tolerance value between variables is the same?
Are you asking when things are multicollinear? Generally that is a correlation of .9 but in regression you can start to have problems around .7.
@@StatisticsofDOOM I have 3 IV’s let’s call them A, B, and C. A and C are showing a correlation of -.584 and their tolerance value is .654 and .650 respectively. The tolerance value for B is .951, so my guess is A and C are showing multicollinearity. Hence, I dropped variable C and re-ran the regression with A and B IV’s and found that their tolerance level was same so, I am interested in know when the tolerance levels are same, are variables showing multicollinearity. Thanks!
@@hinzamalik4467 It's actually very low tolerance values that are problematic, so nothing you've described sounds like multicollinearity.
Thanks for clarifying that so, I should still keep all three variables even though A and C are significantly correlated?
@@hinzamalik4467 If that's what you proposed to do in the analysis.
Hi, why do I get h0 and h1 in my multiple regression results?
H0 is the model with no predictors and just the intercept, while H1 is the model with the predictors. You want the model with the predictors H1 to be better than a model with no predictors H0.
@@StatisticsofDOOM Ah I see, thank you for the clarification 🙏🏼