Good Information for testing the assumptions for ANCOVA in SPSS. It helped me understand my research currently in progress if I ever need to use this information again, I know where to find it. Thank you.
Thank you very much, really helpful. However, I read that when testing the assumption of homogeneity of regression slopes, you need to include the dependent variable and the covariate in the model, too, not just their interaction (like it is done at 08:30 of the video). Otherwise, they say the variance in the DV may be attributed to the interaction term when it shouldn't. This happened in my case and caused a panic attack and desperate web-searches :) No one seems to say clearly how to amend the violation of that assumption. The information I cite is taken from section 16.2 of this book: Field, A. (2013). Discovering statistics using IBM SPSS Statistics (4th ed.). Thousand Oaks, CA: SAGE Publications.
Yes I had the exact same issue after watching this video. My data met all the assumptions except this one, and it took me a while to figure out what I had done wrong. Thanks so much for the video, but maybe edit that bit? :)
Not sure if this will help you (but maybe someone else!) but I think if the homogeneity of regression slopes assumption is violated, you should not proceed with conducting ANCOVA, and should instead conduct multilevel linear modelling. My lecturer was confused by that text (Field, 2013), because in his example, the regression slopes assumption is violated, and yet he continues to conduct ANCOVA anyway, when one should really be doing multilevel linear modelling instead.
Dr. Grande, When testing for homogeneity of regression slopes in the customized model you should have included the main effects of the IV and DV as well as their interaction. - Dr. Napoli
Hi, thank you very much Dr. Grande for your informative videos, they help me a lot for my master's thesis. I would like you to ask you what to do when some assumptions for ANCOVA were not met. I have 3 groups with different types of instructions and their effects on reading comprehension with a pretest-posttest design. The assumption of linearity was not met because while the post-test results of the experimental groups increased, it decreased for the control group which had a significantly higher mean in the pretest. Also, the homogeneity of regression shows a statistically significant interaction effect (p=.032). When the assumptions were not met, the use of ANCOVA was not recommended, however, I cannot figure out an alternative inferential statistics for it despite days of research. Should I just continue with ANCOVA (would it affect the results or result in type I or II errors?) or is there any alternative to overcome these violations? Sorry for the long message and taking your time. I would appreciate any suggestions. Thank you.
Hi Dr. Grande, thank you for very educative videos. I have one question, what should I do when the assumptions of homogeneity of both regression slopes and variance are violated?
Hi Todd, I'm got a question. Hopefully you see this and can answer! In your video "ANCOVA in SPSS" when testing for homogeneity of slopes you use the full model, with the covariate and independent variable as well as their interaction included. However, you only include the interaction term in this video. Which would you use to determine if slopes are different reliably, the one with the full model?
Bruh u saved my life had an assignment due tomorrow and I couldn't figure out how to do the last bit. Thanks Dr. Grande
I really appreciated you pointing out the differences between ANCOVA and ANOVA there at the end- that really pulled it altogether for me conceptually
Excellent. Agree with Corie....comparison of ANCOVA and ANOVA immensely helpful in conceptualizing this.
Thank you Dr. Grande. This VDO is very helpful for my project.
what happend if my datas violate the Homogeneity of Regression Slopes?
Good Information for testing the assumptions for ANCOVA in SPSS. It helped me understand my research currently in progress if I ever need to use this information again, I know where to find it. Thank you.
You saved my ass, man. Thank you so much for this. I actually understood what you were doing and what you were talking about. THANK YOU!
Very helpful for my master's thesis, thank you!!!
That was an excellent explanation. Just have reservations about the violations of linearity.
Excellent explanation. Thank you so much for being clear and precise. This helps me a lot
You're welcome - thanks for watching -
Thank you very much, really helpful.
However, I read that when testing the assumption of homogeneity of regression slopes, you need to include the dependent variable and the covariate in the model, too, not just their interaction (like it is done at 08:30 of the video). Otherwise, they say the variance in the DV may be attributed to the interaction term when it shouldn't.
This happened in my case and caused a panic attack and desperate web-searches :) No one seems to say clearly how to amend the violation of that assumption.
The information I cite is taken from section 16.2 of this book:
Field, A. (2013). Discovering statistics using IBM SPSS Statistics (4th ed.). Thousand Oaks, CA: SAGE Publications.
Yes I had the exact same issue after watching this video. My data met all the assumptions except this one, and it took me a while to figure out what I had done wrong. Thanks so much for the video, but maybe edit that bit? :)
Not sure if this will help you (but maybe someone else!) but I think if the homogeneity of regression slopes assumption is violated, you should not proceed with conducting ANCOVA, and should instead conduct multilevel linear modelling. My lecturer was confused by that text (Field, 2013), because in his example, the regression slopes assumption is violated, and yet he continues to conduct ANCOVA anyway, when one should really be doing multilevel linear modelling instead.
Dr. Grande, When testing for homogeneity of regression slopes in the customized model you should have included the main effects of the IV and DV as well as their interaction. - Dr. Napoli
That was my impression as well...
Excellent description of the assumptions, thanks.
Hi, thank you very much Dr. Grande for your informative videos, they help me a lot for my master's thesis. I would like you to ask you what to do when some assumptions for ANCOVA were not met. I have 3 groups with different types of instructions and their effects on reading comprehension with a pretest-posttest design. The assumption of linearity was not met because while the post-test results of the experimental groups increased, it decreased for the control group which had a significantly higher mean in the pretest. Also, the homogeneity of regression shows a statistically significant interaction effect (p=.032). When the assumptions were not met, the use of ANCOVA was not recommended, however, I cannot figure out an alternative inferential statistics for it despite days of research. Should I just continue with ANCOVA (would it affect the results or result in type I or II errors?) or is there any alternative to overcome these violations? Sorry for the long message and taking your time. I would appreciate any suggestions.
Thank you.
What if there’s outliers. What should I do?
What do I do if I have more than one dependent variable the line only takes one variable
Hi Dr. Grande, thank you for very educative videos. I have one question, what should I do when the assumptions of homogeneity of both regression slopes and variance are violated?
Thank you! what about the Homoscedasticity assumption?
How to handle the violation of homogeneity of variances?
do we have to write and show these assumptions in our research paper? or is it ok to just show ancova in paper?
Hi Todd, I'm got a question. Hopefully you see this and can answer! In your video "ANCOVA in SPSS" when testing for homogeneity of slopes you use the full model, with the covariate and independent variable as well as their interaction included. However, you only include the interaction term in this video. Which would you use to determine if slopes are different reliably, the one with the full model?
I was wondering what you would recommend if your data was not normally distributed and you had a lot of outliers?
Transform it
thank you dr. grande
Very helpful thank you!!
Thank you!
yes, very helpful.