Funny that the example was sleep deprivation, because I fell asleep during the lectures on main effects and interactions and missed learning this. Thank you so much for helping me out!
I hope you all find this video useful! If you have any questions, please do drop me an email. And if the video has REALLY helped and you want to buy me a beer, please see www.paypal.com/donate/?hosted_button_id=L692FVF8PNUKU
After a day of trying to figure this out in my textbook, I finally looked it up on RUclips. You just resolved a days worth of stress. Fuck the textbook. I’m using RUclips from now on.
Its pretty helpful, but when explaining about finding the averages between the 2 points and seeing if they are main effects, you never actually say if they are main effects or not. Or if the averages are indeed different, so that they are main effects.
what was the value of the interaction? I understand that the change in the caffeine pill sample was small, 10, and the change in the placebo was great, 75, so does this make the interaction effect 65? thanks :)
After reading text books , reading so many articles and research paper and seeing many youtube videos I thought of giving up the topic but at last I found your video and it made the concept all clear. Thanks a lot from India for uploading such nice video on this complex topic.
Hi Jim, thank you for your simplified version of understanding fANOVA! Choosing the right statistical test gets quite difficult with many groups, and multi-dimensional output variables, this helped a lot!
What a great video! I appreciate it so much, I wonder if you can go into describing the kind of interactions there are and the qualitative and quantitative aspects of an interaction? I see that they are diverging but it's hard for me to understand what that means. Would this be a quantitative interaction because caffeine strengthens the other Independent variable? (Behavioral Stats Exam coming up this week) Thank you again!
thank you SOOOO MUCH. I have a single course exam tomorrow, if i pass im gonna graduate. If i had money i would definitely buy a bottle of fine champagne but i dont :( I HOPE I PASS TOMORROW
Amazing video! Very well explained, thank you. Do you have another video where you explain how to test whether these effects are significant or not? Would be greatly appreciated!
Thank you. I hink you only forgot to mention that if they will eventually INTERSECT on one end of the lines ( which in this case yes they will intersect on the upper left) meaning if we used a different amount of caffeine in the pill or had them skeep deprived less than 1 hour. No intersection = no interaction.
So in conclusion if we take caffein supplements we can continuously deprive ourselves from sleep and still perform decently! NOTED. I will never sleep again.
I learned how to look up F scores in the table, and that the root of an F-score equals a T-score. But what happens with the Degrees of freedom? how do you know which one to use to use the T-table?
Hi Kason (I can't reply directly as you've used a linked comment!) - yes, that's a great way of wording the interaction! Bear in mind it doesn't describe *what* that moderation was (i.e. did it make the effect of sleep deprivation get worse, or better?), so you might need to add a qualifying statement. Good stuff, though!
A good way to describe the interaction in words: "The type of supplement given (caffeine or placebo) moderated the effects of sleep deprivation on memory recall (or whatever other dependent variable)." Correct me if I am wrong. I am in the process of trying to understand how to describe interactions in words, haha.
Thank you for this video!!! I was tied up in knots over my main effects and interactions as there was significant main effects but not interactions. Made perfect sense. Funny how you can spend 3years at uni being lost in statistics, yet a simple video works wonders :-D Thank you!
Does the same explanation apply to a moderated regression analysis ? Let's say we get all coefficients as significant / moderator is dummy coded 1 for women and 0 for men. If we have a significant negative coefficient for the interaction term, we could say there is a decrease in Y as X increases, but this decrease is stronger for the women's group? How can we better explain the interaction effect when reporting results in a journal? The effect of X on Y is stronger in women? Do we report the difference in numbers? Could you provide an example of how to formally report a moderated regression with a dummy variable?
Hi :) thank you very much for your helpfull videos!! i have a question about second figure. i could not be sure if it is ordinal or disordinal... Because in my figure one group is crossing worh other 2 groups but these two groups are not crossing with each other but they are getting closer to each other. What my interpretation should be? Ordinal, disordinal or shoul i interpret seperately? :/
Do I understand correctly that the effects of the single variables are negative (negative slope) whilst the interaction effect is positive (alleviates the negative effect of sleep deprivation)?
the vid my professor attached to teach us this cuz it's an online class is 2 hours long. You just summed up 2 hours, held my attention, and made it make sense. omg THANK YOU!!!!!! - a sleep deprived nursing student who's trying to self teach herself stats
If the main effects are significant but the interaction effect is not, is it acceptable to continue with a posthoc test for multiple comparisons, or is the posthoc test reserved for when an interaction effect exists?
Thanks Jim for the video. This is slightly off-topic but I was wondering what your thoughts were on this issue. Do you think it's a bad idea to standardize (z-transform) already Box-Cox transformed dependent variables for multivariate scaling purposes prior to a MANOVA? For my dataset, BC did successfully pick the right lambda and "normalize" the data i.e. residuals follow gaussian distribution now. However, it didn't do much to "standardize" the scale. For instance, Min-max for one of the DVs is on a scale of 1000-2000 whereas for another DV, on a scale of 10-40. So, I thought post-BC standardization of scale would be necessary to de-emphasize impact of higher variance from 1st DV on the second in a composite. I am concerned if reviewers will frown upon Z-transforming a dataset that's already BC transformed because of some statistical reasoning or the way these two transformations when done together alter the dataset that I am not aware of.
+scientistUMT Hi - to be honest, I am not sure. Do you get conflicting results with transformed and non-transformed analysis? Reviewers may wish to see alternative analysis, so if you have already done this and it doesn't contradict your main findings then this will be OK.
Excellent video. Searching on RUclips for the last hour or so for a clear explanation on the meaning of main effects and interaction and this video was the best by far.
the next step would be the POSTHOC analysis. if the interaction term is significant, the two groups should be analyzed separately with one-way ANOVA (y=a for b=1, y=a for b=2). if not, consider a model without the interaction term, but that has two covariates ( y= a+b+error).
Funny that the example was sleep deprivation, because I fell asleep during the lectures on main effects and interactions and missed learning this. Thank you so much for helping me out!
Hahaha, perfect :)
lmao same i just didnt show up to lecture
I go to school in the U.S. and was so confused learning about this, but this video helped me understand factorial designs better, thank you!
thank you - just completing the last assignment of my Research Methods in Education / online and this helped a lot Jim ! Cheers fro NZ :D
Thank you so much!!! I was really struggling with this concept and not being able to wrap my head around it.
I'm glad it helped!
Thank you very much
Helped a lot!
great explanation ,thank you
Well done.
amazing explanation, thanks boss
WISH YOU WERE MY LECTURER
Thanks!
Oh my goodness! I finally get it! Thank you so much for making this topic so understandable.
I hope you all find this video useful! If you have any questions, please do drop me an email. And if the video has REALLY helped and you want to buy me a beer, please see www.paypal.com/donate/?hosted_button_id=L692FVF8PNUKU
really good!Thank you very much!
THANK YOU SO SO SO MUCH!
This was so helpful and clear. I have struggled with this my entire undergraduate year.
Great explanation! Thank you.
Hi Jim, I didn't sleep well, I'm gonna take a big dose of caffeine, thanks for the advice :) and thank you a lot for the explanations.
Was going to send a few bob, but PayPal fee would have got as much as you! So sorry but thanks anyway
I'm not even from this uni and I wish you were one of my lectures 😔
After a day of trying to figure this out in my textbook, I finally looked it up on RUclips. You just resolved a days worth of stress. Fuck the textbook. I’m using RUclips from now on.
Its pretty helpful, but when explaining about finding the averages between the 2 points and seeing if they are main effects, you never actually say if they are main effects or not. Or if the averages are indeed different, so that they are main effects.
Actually at the end he's doing so - there are main effects for both variables :)
what was the value of the interaction? I understand that the change in the caffeine pill sample was small, 10, and the change in the placebo was great, 75, so does this make the interaction effect 65?
thanks :)
this was soo clear :)
After reading text books , reading so many articles and research paper and seeing many youtube videos I thought of giving up the topic but at last I found your video and it made the concept all clear. Thanks a lot from India for uploading such nice video on this complex topic.
The easiest to understand explanation of all the different videos i have looked at. Perfect! Thank you!
Hi Jim, thank you for your simplified version of understanding fANOVA!
Choosing the right statistical test gets quite difficult with many groups, and multi-dimensional output variables, this helped a lot!
What a great video! I appreciate it so much, I wonder if you can go into describing the kind of interactions there are and the qualitative and quantitative aspects of an interaction? I see that they are diverging but it's hard for me to understand what that means. Would this be a quantitative interaction because caffeine strengthens the other Independent variable? (Behavioral Stats Exam coming up this week) Thank you again!
thank you SOOOO MUCH. I have a single course exam tomorrow, if i pass im gonna graduate. If i had money i would definitely buy a bottle of fine champagne but i dont :( I HOPE I PASS TOMORROW
Amazing video! Very well explained, thank you. Do you have another video where you explain how to test whether these effects are significant or not? Would be greatly appreciated!
im able to pay attention way better w the accent than if i were litening to an american professor
Ha, thank you.
AMAZING VIDEO. you're saving lives, my friend!
Thank you! Such a great and clear explanation!
Thank you! This might have saved me
amazing explanation! THANK YOU
Most helpful. The best explanation I've seen on this. Thank you.
Thank you so much!!!! .You save my final examination's score!!!
literally same
Küsse deine Augen, Bruder
Thank you. I hink you only forgot to mention that if they will eventually INTERSECT on one end of the lines ( which in this case yes they will intersect on the upper left) meaning if we used a different amount of caffeine in the pill or had them skeep deprived less than 1 hour. No intersection = no interaction.
AMAZING! Thank you so much for this!
This is a super helpful explanation to a relatively complex topic, it makes it seem very intuitive and I wanted to thank the author of this video
Glad it helped! Thanks for watching.
this makes so much more sense now, thanks!
Thank you from the University of Cape Town!
So in conclusion if we take caffein supplements we can continuously deprive ourselves from sleep and still perform decently! NOTED. I will never sleep again.
Ha, well, we need data on that particular experimental setup! :)
Very well explained i would imagne the math across multiple dimensions might be a bit more complex
I learned how to look up F scores in the table, and that the root of an F-score equals a T-score. But what happens with the Degrees of freedom? how do you know which one to use to use the T-table?
Hi Kason (I can't reply directly as you've used a linked comment!) - yes, that's a great way of wording the interaction! Bear in mind it doesn't describe *what* that moderation was (i.e. did it make the effect of sleep deprivation get worse, or better?), so you might need to add a qualifying statement. Good stuff, though!
A good way to describe the interaction in words: "The type of supplement given (caffeine or placebo) moderated the effects of sleep deprivation on memory recall (or whatever other dependent variable)." Correct me if I am wrong.
I am in the process of trying to understand how to describe interactions in words, haha.
I have an exam tomorrow for my research methods class, thank you for this video!
This may sound dumb but couldn't we just look at the slope for each line?
Thanks! You really helped me understand this better.
Clear explanation. This helps me a lot! Thank you~~
Oh my GOODNESS thank you so much for this. My discussion went over it so quickly I didn't catch any of it! Thank you a thousand times over
best explanation on this topic .Thank you.
I never would have understood this reading the textbook
can we just say there is an interaction when slopes are different?
Great explanation. Thank you.
my professor uses bar graphs for this and it makes way less sense. this is much easier to look at to me lol
I'm not in your class but this is going to help me pass mine! Thanks Jim!
Thanks for the kind words!
this helped me so much!! thank you
Thank you for this video!!! I was tied up in knots over my main effects and interactions as there was significant main effects but not interactions. Made perfect sense. Funny how you can spend 3years at uni being lost in statistics, yet a simple video works wonders :-D Thank you!
thanks for the simple clarification
Thank you sir for making sense of this topic :)
What certain test to perform if the interaction is significant?
「上記のギフトのいずれかを選択できます」、
This was super useful! Thanks a bunch!
Does the same explanation apply to a moderated regression analysis ? Let's say we get all coefficients as significant / moderator is dummy coded 1 for women and 0 for men. If we have a significant negative coefficient for the interaction term, we could say there is a decrease in Y as X increases, but this decrease is stronger for the women's group? How can we better explain the interaction effect when reporting results in a journal? The effect of X on Y is stronger in women? Do we report the difference in numbers? Could you provide an example of how to formally report a moderated regression with a dummy variable?
Yes, a moderated regression is the same as an interaction. ANOVA is basically regression in disguise.
THANK YOU THANK YOU THANK YOUUUUUUUUUUU!!! Sorry for the caps, but I've never seen SUCH A CLEAR explanation of this. THANK YOU!
Nice examples. Thanks.
best video on the subject
Well explained, thanks!
Thank you so much Jim!
Very helpful video. Thank you for taking the time to post it and make it public.
My pleasure! Thanks for your feedback
Thank you very much for your explanation!!! This video really helps me understand better about main effects and interactions!!!
Nice illustration mate!
Very clear !!! Thanks 😊
This sleep deprived Tilburg University student thanks you very much! Great explanation and very useful for my research note.
Hi :) thank you very much for your helpfull videos!! i have a question about second figure. i could not be sure if it is ordinal or disordinal... Because in my figure one group is crossing worh other 2 groups but these two groups are not crossing with each other but they are getting closer to each other. What my interpretation should be? Ordinal, disordinal or shoul i interpret seperately? :/
Hello Sir, I have to draw an interaction graph where gender is my moderator. Could you please elaborate how to PLOT such interaction graph?
Thank you so much Jim. That really made it clear. You have a superb way of explaining things.
Really good explaining!
Hmm.. what if there is just one factor. I don’t believe there would be an interaction but I assume there would be a main effect ??
That's correct - you can only have interactions when you have more than one factor.
Best explanation ever
Do I understand correctly that the effects of the single variables are negative (negative slope) whilst the interaction effect is positive (alleviates the negative effect of sleep deprivation)?
Thank you so much, this really explained things clearly.
the vid my professor attached to teach us this cuz it's an online class is 2 hours long. You just summed up 2 hours, held my attention, and made it make sense. omg THANK YOU!!!!!! - a sleep deprived nursing student who's trying to self teach herself stats
Really great vid!!!
excellent. well simplified. i have clearly understood
macam mana nak buat?
Okay good video but you never actually said anything about the results of your main effect analysis.
The video is about conceptual understanding of what a main effect *is* rather than giving you p-values.
can you please explain why do we always assume that there is no interaction effect between blocks and treatments in a randomized block design?
This is the video that helped my brain wrap around the topic main effect and interaction. Thank very much !
Thank you this is so much more clear than the mess my lecturer left me with!!
Very simple to understand the complex concept, thank you very much
If the main effects are significant but the interaction effect is not, is it acceptable to continue with a posthoc test for multiple comparisons, or is the posthoc test reserved for when an interaction effect exists?
thank u Jim.
A student from Greenwich university, this was VERY helpful!
My pleasure!
i got a 2:1 in my exam thank u :)
Superb! Congratulations! :)
Thank you,it’s very useful!
Very clear explanation--thanks, sir.
Thanks Jim for the video. This is slightly off-topic but I was wondering what your thoughts were on this issue. Do you think it's a bad idea to standardize (z-transform) already Box-Cox transformed dependent variables for multivariate scaling purposes prior to a MANOVA? For my dataset, BC did successfully pick the right lambda and "normalize" the data i.e. residuals follow gaussian distribution now. However, it didn't do much to "standardize" the scale. For instance, Min-max for one of the DVs is on a scale of 1000-2000 whereas for another DV, on a scale of 10-40. So, I thought post-BC standardization of scale would be necessary to de-emphasize impact of higher variance from 1st DV on the second in a composite. I am concerned if reviewers will frown upon Z-transforming a dataset that's already BC transformed because of some statistical reasoning or the way these two transformations when done together alter the dataset that I am not aware of.
+scientistUMT Hi - to be honest, I am not sure. Do you get conflicting results with transformed and non-transformed analysis? Reviewers may wish to see alternative analysis, so if you have already done this and it doesn't contradict your main findings then this will be OK.
Most useful exploration of this I have read or seen - thank you
you are great
Tx! I ll watch it more times but I do understand it.
So clear to understand, thanks you so much!!!
In order to be really sure whether or not there were significant differences in the data what should be included in the graphs that is missing?
Excellent video. Searching on RUclips for the last hour or so for a clear explanation on the meaning of main effects and interaction and this video was the best by far.
Thank you sooooo much for this video!! My GPA has been saved.
can I have clarity about further analysis after getting interaction effect significant in 2*2 factorial design...
the next step would be the POSTHOC analysis.
if the interaction term is significant, the two groups should be analyzed separately with one-way ANOVA (y=a for b=1, y=a for b=2).
if not, consider a model without the interaction term, but that has two covariates ( y= a+b+error).