Hi I appreciate your valuable information! I have a question about this. Could you tell me why you chose "safe_risky" as a variable to split the data set? That is, I don't understand why you isolateed the interaction between "like_dislike" and "self_other" across "safe_risky". I would appreciate your response!
Hello. Thanks for the explanation. My question is what to do when the sum of squares and the degree of freedom in one of the three two-way interactions and the three-way interaction equal zero. Thank you!
Hello, thank you for the video! I was wondering if you would recommend p-value adjustments when following up on a significant three-way interaction with 2 two-way ANOVAS in a 2x2x2 between-subjects design? Or if you would be able to suggest any references on this matter?
This is a controversial topic, with some recent discussion here: link.springer.com/article/10.3758/s13423-015-0913-5 In my opinion, in the case of an omnibus ANOVA interaction effect that is found to be significant via the F-test, you can follow it up with three 2x2 ANOVAs, without inflating the alpha rate. So, no need to apply any adjustments, in that case. I base this opinion on the simulation research that has shown the Fisher's protected LSD to be good at keeping alpha near .05, when there are three means compared in the omnibus ANOVA. I discuss this in the how2statsbook.com chapter 7.
From memory, David Howell's textbook has such an example. If I came across an interesting example of such an analysis (and I could simulate the data sufficiently accurately), I'd include in my textbook. What is your example? Do you think you have a statistically significant three-way interaction?
@@how2statsbook477 Thank you. I have a 2 (treatment group: a and b) x 2 (treatment preference: a and b) x 2 (time: baseline and post) mixed design with time as the within subjects variable. Dependent variable is test scores. It looks like it is a sig. three-way interaction.
Hello! I have conducted a 3 way factorial ANOVA and violated the Levenes test. There was also a problem when gathering my data so I have a huge variance in participant numbers between groups. Is there anything relatively simple that could be done to sort this? Thanks
For factorial ANOVAs, it can be very difficult to interpret the results, when the sample sizes are substantially unequal. Check out Howell's discussion: www.uvm.edu/~statdhtx/StatPages/More_Stuff/Unequal-ns/unequal-ns.html In the first instance, I would run the analysis with your existing data and estimate the power (there's a button option for that in SPSS). Then, randomly sample from each of your groups the number of cases you have equal to the smallest cell size. Re-run the analysis with that smaller set of data and estimate power. How much do you lose?
I have question! All my degree of freedom goes into the error but not into the interaction part but when I did the experiment I did the interaction part. Is there any problem of the way I key in the data?
It's really tough for me to answer that question. However, if you want to explore how the data are set-up for this analysis and practice with the SPSS data file, you can get all of the how2statsbook data files here: www.how2statsbook.com/p/data-files.html
Hi, that is really appreciated!. Amazing interpretation. What if I have more than three independent variables (i.e. 6) and three of them have two levels and the other three have more than two levels. In this case, what should I do please? should I use bonferroni too?.
With 6 independent variables, you would be potentially looking at a six-way interaction, which is essentially impossible to interpret. You can either look at the independent variables individually, or, if you're ambitious, conduct a multiple regression with your six independent variables and one dependent variables. You'll have to use effect coding for your independent variables with more than three levels (assuming they are measured on a nominal scale). I don't have a video on that though. Perhaps someone else does?
how2statsbook much appreciation for the reply! I now understand the issue. I think in this case, Rstudio is much more stronger in dealing with interactions compared to SPSS? I have not used it at all but not sure if it would help?
Hi there, What do you do when you have violated Lavenes test in a 3 way mixed design anova using univariate glm in spss? Would you interpret from the Corrected model in between subjects analysis output?
Just to note I have violated Lavenes test with 2 groups one 33 participants and one 32 participants. Is this difference between groups problematic enough to assume that I can't trust the robustness of my p value?
I'd say you don't have a problem, on the basis of my review of the area (discussed in my textbook, Chapter 11). Also, if the larger variance (standard deviation) is associated with the larger group (33 participants), the statistical analysis will actually be slightly conservative.
Hi there great video! I also have some questions. I'm trying to conduct an experiment with 3 independent variables but im not sure about the test i should use, the one you presented its between subject Mine I think its mixed design with between and within. what test should i do? for parametric and non parametric. independent variables are: gender- male, female ; time- 5 days, 10 days; interview technique- sketch, mrc. dependent variabile- details type- person, action, location
Hi Sir, I have a question. For my thesis I have a similar design however my dependent variable is attitude and measured on 1-7 Likert scale. Could you run the same analysis in this case? Kind regards and hope you have the time to answer.
Once again, your vid has saved me from being depressed over stats. Big thank you. This is very helpful
Hi I appreciate your valuable information!
I have a question about this. Could you tell me why you chose "safe_risky" as a variable to split the data set? That is, I don't understand why you isolateed the interaction between "like_dislike" and "self_other" across "safe_risky". I would appreciate your response!
Hello. Thanks for the explanation. My question is what to do when the sum of squares and the degree of freedom in one of the three two-way interactions and the three-way interaction equal zero. Thank you!
How to calculate one way anova if under heads subheads made calculation of head including subheads for anova test
Thx for your helpful video!
Can we perform Post Hoc Mean Separation test in between main factors?
Thank you very much
Thank you 👍🏻
This was very helpful! Thank you!
how do you determine the g power in a three way anova?
Hello, thank you for the video! I was wondering if you would recommend p-value adjustments when following up on a significant three-way interaction with 2 two-way ANOVAS in a 2x2x2 between-subjects design? Or if you would be able to suggest any references on this matter?
This is a controversial topic, with some recent discussion here: link.springer.com/article/10.3758/s13423-015-0913-5 In my opinion, in the case of an omnibus ANOVA interaction effect that is found to be significant via the F-test, you can follow it up with three 2x2 ANOVAs, without inflating the alpha rate. So, no need to apply any adjustments, in that case. I base this opinion on the simulation research that has shown the Fisher's protected LSD to be good at keeping alpha near .05, when there are three means compared in the omnibus ANOVA. I discuss this in the how2statsbook.com chapter 7.
@@how2statsbook477 Thank you very much for your reply!
Thank you so much! Super helpful :)
Hi, do you know a good reference for running a 2x2x2 Anova with 1 within-subjects factor (time)?
From memory, David Howell's textbook has such an example. If I came across an interesting example of such an analysis (and I could simulate the data sufficiently accurately), I'd include in my textbook. What is your example? Do you think you have a statistically significant three-way interaction?
@@how2statsbook477 Thank you. I have a 2 (treatment group: a and b) x 2 (treatment preference: a and b) x 2 (time: baseline and post) mixed design with time as the within subjects variable. Dependent variable is test scores. It looks like it is a sig. three-way interaction.
Hello! I have conducted a 3 way factorial ANOVA and violated the Levenes test. There was also a problem when gathering my data so I have a huge variance in participant numbers between groups. Is there anything relatively simple that could be done to sort this? Thanks
For factorial ANOVAs, it can be very difficult to interpret the results, when the sample sizes are substantially unequal. Check out Howell's discussion: www.uvm.edu/~statdhtx/StatPages/More_Stuff/Unequal-ns/unequal-ns.html
In the first instance, I would run the analysis with your existing data and estimate the power (there's a button option for that in SPSS). Then, randomly sample from each of your groups the number of cases you have equal to the smallest cell size. Re-run the analysis with that smaller set of data and estimate power. How much do you lose?
I have question! All my degree of freedom goes into the error but not into the interaction part but when I did the experiment I did the interaction part. Is there any problem of the way I key in the data?
It's really tough for me to answer that question. However, if you want to explore how the data are set-up for this analysis and practice with the SPSS data file, you can get all of the how2statsbook data files here:
www.how2statsbook.com/p/data-files.html
Hi, that is really appreciated!. Amazing interpretation. What if I have more than three independent variables (i.e. 6) and three of them have two levels and the other three have more than two levels. In this case, what should I do please? should I use bonferroni too?.
With 6 independent variables, you would be potentially looking at a six-way interaction, which is essentially impossible to interpret. You can either look at the independent variables individually, or, if you're ambitious, conduct a multiple regression with your six independent variables and one dependent variables. You'll have to use effect coding for your independent variables with more than three levels (assuming they are measured on a nominal scale). I don't have a video on that though. Perhaps someone else does?
how2statsbook much appreciation for the reply! I now understand the issue. I think in this case, Rstudio is much more stronger in dealing with interactions compared to SPSS? I have not used it at all but not sure if it would help?
Hi there,
What do you do when you have violated Lavenes test in a 3 way mixed design anova using univariate glm in spss? Would you interpret from the Corrected model in between subjects analysis output?
Are the sample sizes equal across the three groups?
Just to note I have violated Lavenes test with 2 groups one 33 participants and one 32 participants. Is this difference between groups problematic enough to assume that I can't trust the robustness of my p value?
@@how2statsbook477 Hi there! Yes.
I'd say you don't have a problem, on the basis of my review of the area (discussed in my textbook, Chapter 11). Also, if the larger variance (standard deviation) is associated with the larger group (33 participants), the statistical analysis will actually be slightly conservative.
@@how2statsbook477 Thank you so very much!
Hi there great video!
I also have some questions. I'm trying to conduct an experiment with 3 independent variables but im not sure about the test i should use, the one you presented its between subject
Mine I think its mixed design with between and within. what test should i do? for parametric and non parametric.
independent variables are: gender- male, female ; time- 5 days, 10 days; interview technique- sketch, mrc.
dependent variabile- details type- person, action, location
Hi Sir, I have a question. For my thesis I have a similar design however my dependent variable is attitude and measured on 1-7 Likert scale. Could you run the same analysis in this case? Kind regards and hope you have the time to answer.
Hi..., what version of spss above?
think he said 25