This video was SO helpful! It has been a couple years since I've done a MANOVA and my graduate mentor wanted me to do one for my thesis, so I really needed the refresher. Thank you for this informative tutorial, Dr. Grande.
I originally watched this video on my lunch break at work, naturally I had to watch it again every lunch break for three days! There are so many assumptions and a lot of information. I like that we learned to conduct the MANOVA with two dependent variables AND interpret it.
Very informative and a lot to take in at the same time, but overall very helpful. I have a better understanding of MANOVA after watching this and was able to follow, for the most part. Will watch again in the future.
Thanks for all your videos. It helped me in completing my thesis. Clear and well structured explanation plus the interpretation. May God bless you always.
Dr. Grande. Please be my multivariate stats teacher! Gosh, it is so clear and easy to understand and I can use spss now to do manova (in our class, we now use SAS)
The Residual's Statistics table has a lot of information, it is almost overwhelming. I am glad that we are currently just looking at the Mahalanobis Distance.
This helped me a lot Dr. Grande. Thank you so much! Its always easier to learn things and find the corresponding literature after than the other way around.
The walk through of the results was good and I followed along for the most part. I am finally understanding the SPSS outputs. I still don't quite follow why tests are picked and how to decide that.
Wow!! that's a lot of Assumptions. I was always taught that making assumptions makes an @ out of U and Me....HA Statistics jokes...how have we not used this one yet? This is great to watch before completing the homework. Dr. Grande said at the beginning of the semester that stats are a series of steps, and if you follow the steps you are OK. it is fascinating how each of these tests is building on the one before it. I am thankful for learning it in this step by step process
Thank you so much for this video, Dr. Todd! It's a very big help (first time student's experiment took on MANOVA)... May I ask, though, if the Levene's Test of Equality of Error Variances is significant, is there a robust test to take on this concern? Thanks!
Yes doctor Grande, I found the video very useful, but at the same time difficult since it's a lot of information to take. I'm sure I'll be watching it again...
Tons of information! I know we are looking for a linear relationship between each pair of dependent variables across each level of independent variables. So then we look at statistically significant variables. Still unsure of why we add the Mahalanobis distance?
Hello Dr Grande, Thanks for the video on MANOVA. I have used MANOVA for my analysis (1 independent variable with 6 groups and 5 dependent variables. The result of my multivariate test for the Pillai's trace are as follows; Value=1.665, F (25, 120)=2.39, p=0.001; partial eta (Ƞ2 =0.333), observed power =0.998. But I am worried about the Pillai's trace value because I assume this should be between 0-1 but mine is greater than 1. Could you help of there is an explanation for this?
Please help your viewers to understand why you reversed the independent and dependent variables when running the test for Mahalanobis. I'm completing my dissertation and need the logic that this test produces a valid value. Thank you so very much! Also, do you have any citation(s) suggested to support removing an outlier that is above the maximum allowable critical Mahalanobis value?
Hello Dr. Grande, I'm new in statistics. Recently we have a statistics report, and I have to do MANOVA. My independent variable, the group, has two levels, while i have three outcomes: "BMI", "endurance", and "PEI". I have checked all of the assumptions in MANOVA, and already exclude those outcomes which violated the assumptions by using other analysis. After running MANOVA, it is ok with the box's test (p = 0.073,> 0.05), multivariate test (all sig. shows 0.019, < 0.05) and Levene's test (all sig. > 0.05). But at the "tests of between-subjects effects" part, in the "group" row, i found all outcomes (dependent variables) shows > 0.05 in sig. column, which means there are no significant difference between groups (if i'm not wrong?). Isn't it should have at least one outcome shows significant difference, since the multivariate test shows < 0.05. So, I would like to ask if there are any possibilities for this situation to happen? Or if not, what should I do for the next step? Thank you very much! Sincerely, Xin Yun, greetings from Taiwan
Thank you for this tutorial.. I have question sir. If the kolmogorov or Shapiro value less than 0.05 , it means the dara not normally distribution,, what we should do in such cases please?
Is it typical to seek both a linear contrast (which is based on prior hypothesis) and a post hoc test that is ordered after the means have been observed ?
Thanks a bunch Dr. Grande. Actually, I wonder if you could help me with my data. I've pre-post treatment for anxiety. My data is collected from a skills test and anxiety scale. Which mixed design test should I use? Anova, Ancova or Manova to identify the differences between the subgroups: 2 (exp vs control group) by 2 (high anxiety vs no anxiety) by 2 (pre- vs post-test)? ( spss has shown that 2 groups/4 are not homogeneous) . By the way, the no of participants is not equal: (exp: 22 & 31; Control: 24 &26). Besides, a negative correlation (0.49) exists between the 2 dependent variables. Your help is highly appreciated.
Thank you! Dr. Todd Grande, I have a question that what is the allowable value for Mah. distance if I have 11 dependent variables, and what to do if I have outliers?
Hello Dr Todd May God bless you. Good job fo A great video and explanation. I have A few question. I follow the video and perform the procedure. Here I got, the data test for outiler, normalities are OK but when looking for the multicolinearity the number is greater than 0.8 and 0.9. What should I? Can I still continue to test the MANOVA? I see your data is also not normalized for quantity what should I do IF we meet this situation. Regards Fahrudin
Is there a non-parametric version of a MANOVA. I have two lists independant variables (1-3 days, and 6 different concentrations) and 4 dependant variables (which are not normally distributed). When I looked on the internet I could only find kruskall-wallis test but that is not exactly what I'm looking for. And is there also a test you van do when the dependant variable is ordinal and you have 2 independant variables?
Hi Dr Grande, I have a quick question. The MANOVA is used when you have a single IV and 2 or more DVS. If the DV;s are not obviously related (Let's say, Self-Reported Sleep Quality for one and Sex Satisfaction for 2) is it still possible to run a MANOVA? Thanks
What if one instead of quantity you had accuracy on a task, that was measured as a binary of correct or incorrect (1 | 0). Surely a MANOVA would not be appropriate, as looking at means would not produce sensical results, right? Would you say that to compare that across groups it would be best to split file by condition and run a frequency analysis on the proportion of correct responses? How would you test if those differences were statistically significant?
Hallo there, If I report the item wording and compare on the Item level, do I need to recode my reversed scored items? For running a MANOVA with negative and positive items do I need to recode my negative items? Or do I only recode when I want to analyse the factorial level? Thank you for your help!
Hi Todd,Thank you for this video. Please make a video on how to set up the data on SPSS especially the data from factorial table. I would like to learn how to verify the findings from journal articles if they only give a table of means to work on.Thank you so much!
i tried a similar analysis for the data I have. I would like to ask you one doubt. What is the data type of the column named program that you have assigned? I tried string it didnt work. If I select numeric I cant enter these type of text. Can you please clarify. thanks.
Hi there Todd, thank you for your videos! I have a question to ask, if my dependent variables for this case are all in Likert Scales.. do i have to compute the variables into one construct and then conduct the assumption testing? Thank you!
If there some variables have no linear relationship across group then what should we do? I looked up in many books (Stevens, tabachinik you name it!) But nothing is given. Should we go for seperate ANOVA for each variable how can we control the family-wise error rate??
Hi, Dr. Grande. I was just wondering how did you find out that the individual is more effective than the treatment as usual and not vice versa? Thank you!
can someone tell me if MANOVA gives you the univariate ANOVAS automatically or you have to run them by yourself ?? Because it gave me some pairwise comparisons for my main effects but not for the interaction. Thank you!
What are your thoughts about the different theories for sample size? Also, do we need to interpret normality with caution since we are testing univariate normality instead of multivariate normality?
Sir I have one question When we conducting experimental study with two level of treatment variable and two dependent variable which is a MANOVA study . Sir to test the assumptions of MANOVA which scores considered Pre test score Post test scores Gain score Please sir reply as soon as possible Thanks sir
Hello Dr Grande Iam a master student and iam conducting a research on writing development of EFL students across four grade levels iam rather focusing on syntactic development, so i 'll be using a set of syntactic complexity measures and i 'll evaluate them across proficiency level( grade 1 to 4) . So, basically i've a continous dependent variable ( Syntactic complexity) with 7 levels and a categorical independent variable ( writing proficiency). Apparentlly i should use one way MANOVA but i faced a problem , though there is a difference between the means of all the syntactic complexity measures examined ( the mean difference is above 3) , post hoc tukey test shows p > 0.5 ( ,99) and some of the mean differences that were significant are acctually in a negative direction, does that have to do with the normality of the distribution of the independent variable or with homogenity of variance -covariance? because according to Box's test the null hypothesis is rejected besides the p in shapiro-wilk normality test is 0.00 . Please if you can explain that to me i'll be so grateful as i'll have to submitt my dissertation in few days and i didn't write the result section yet. REGARDS
Hi, did you get your answer yet? I have a similar question and if you got the answer for the same, it would be quite helpful to for me to learn about it
God bless you!your tutorial literally saved my thesis! Thank you very much!
You're welcome - thank you for watching.
This video was SO helpful! It has been a couple years since I've done a MANOVA and my graduate mentor wanted me to do one for my thesis, so I really needed the refresher. Thank you for this informative tutorial, Dr. Grande.
You're welcome!
I originally watched this video on my lunch break at work, naturally I had to watch it again every lunch break for three days! There are so many assumptions and a lot of information. I like that we learned to conduct the MANOVA with two dependent variables AND interpret it.
THANK YOU SOOOO MUCH! I tried watching multiple videos and yours was the only one that helped me!
Very informative and a lot to take in at the same time, but overall very helpful. I have a better understanding of MANOVA after watching this and was able to follow, for the most part. Will watch again in the future.
I find all your videos of statistical procedures in SPSS, extremely helpful.
Thank you very much!
Thanks for all your videos. It helped me in completing my thesis. Clear and well structured explanation plus the interpretation. May God bless you always.
Dr. Grande. Please be my multivariate stats teacher! Gosh, it is so clear and easy to understand and I can use spss now to do manova (in our class, we now use SAS)
The Residual's Statistics table has a lot of information, it is almost overwhelming. I am glad that we are currently just looking at the Mahalanobis Distance.
+Amanda Sutton Yes it does. Where does it all come from?
Incredible explanation where every other tutorials failed. Thanks professor.
This helped me a lot Dr. Grande. Thank you so much! Its always easier to learn things and find the corresponding literature after than the other way around.
Thanks for defining the cutoffs for multicolinearity, that helped quite a bit.
That was a ton of information. Testing the assumptions took as much or more time than doing the actual statistical analysis with the MANOVA.
The walk through of the results was good and I followed along for the most part. I am finally understanding the SPSS outputs. I still don't quite follow why tests are picked and how to decide that.
same
Hello Dr. Grande! I loved your video of conducting MANOVA on SPSS!
If you are interested in conducting MANOVA with R, there is a video on my channel that might be helpful :)
This was a great introduction to Manova. Thanks Dr. Grande!
Wow!! that's a lot of Assumptions. I was always taught that making assumptions makes an @ out of U and Me....HA Statistics jokes...how have we not used this one yet? This is great to watch before completing the homework. Dr. Grande said at the beginning of the semester that stats are a series of steps, and if you follow the steps you are OK. it is fascinating how each of these tests is building on the one before it. I am thankful for learning it in this step by step process
Thank you so much for this video, Dr. Todd! It's a very big help (first time student's experiment took on MANOVA)... May I ask, though, if the Levene's Test of Equality of Error Variances is significant, is there a robust test to take on this concern? Thanks!
Yes doctor Grande, I found the video very useful, but at the same time difficult since it's a lot of information to take. I'm sure I'll be watching it again...
Very informative video! Gotta question: what if one of the assumptions are not met?
If assumptions are not met, you interpret Pillai's Trace from the table. @Todd can correct me if I am wrong?
Tons of information! I know we are looking for a linear relationship between each pair of dependent variables across each level of independent variables. So then we look at statistically significant variables. Still unsure of why we add the Mahalanobis distance?
it will be help if you include a data set also for practice by students.
Very detailed coverage of assumptions and running it. Thanks !
You're welcome. Thanks for watching.
Hello Dr Grande, Thanks for the video on MANOVA. I have used MANOVA for my analysis (1 independent variable with 6 groups and 5 dependent variables. The result of my multivariate test for the Pillai's trace are as follows; Value=1.665, F (25, 120)=2.39, p=0.001; partial eta (Ƞ2 =0.333), observed power =0.998. But I am worried about the Pillai's trace value because I assume this should be between 0-1 but mine is greater than 1. Could you help of there is an explanation for this?
Really helpful! I am using a MANOVA for my thesis. thank you so much!
sir, it is very nice explanation, could you tell what tables we should include in paper writing for journal thanks
Please help your viewers to understand why you reversed the independent and dependent variables when running the test for Mahalanobis. I'm completing my dissertation and need the logic that this test produces a valid value. Thank you so very much! Also, do you have any citation(s) suggested to support removing an outlier that is above the maximum allowable critical Mahalanobis value?
Hello Dr. Grande, I'm new in statistics. Recently we have a statistics report, and I have to do MANOVA. My independent variable, the group, has two levels, while i have three outcomes: "BMI", "endurance", and "PEI".
I have checked all of the assumptions in MANOVA, and already exclude those outcomes which violated the assumptions by using other analysis.
After running MANOVA, it is ok with the box's test (p = 0.073,> 0.05), multivariate test (all sig. shows 0.019, < 0.05) and Levene's test (all sig. > 0.05). But at the "tests of between-subjects effects" part, in the "group" row, i found all outcomes (dependent variables) shows > 0.05 in sig. column, which means there are no significant difference between groups (if i'm not wrong?). Isn't it should have at least one outcome shows significant difference, since the multivariate test shows < 0.05.
So, I would like to ask if there are any possibilities for this situation to happen? Or if not, what should I do for the next step? Thank you very much!
Sincerely, Xin Yun, greetings from Taiwan
You explained this very clear, thank you very much!
You're welcome -
Thank you for this tutorial.. I have question sir.
If the kolmogorov or Shapiro value less than 0.05 , it means the dara not normally distribution,, what we should do in such cases please?
Hi Dr. Grande, what to do when the assumption of linearity is not met for all cases?
Is it typical to seek both a linear contrast (which is based on prior hypothesis) and a post hoc test that is ordered after the means have been observed ?
Thanks a bunch Dr. Grande. Actually, I wonder if you could help me with my data. I've pre-post treatment for anxiety. My data is collected from a skills test and anxiety scale.
Which mixed design test should I use? Anova, Ancova or Manova to identify the differences between the subgroups:
2 (exp vs control group) by 2 (high anxiety vs no anxiety) by 2 (pre- vs post-test)? ( spss has shown that 2 groups/4 are not homogeneous)
.
By the way, the no of participants is not equal: (exp: 22 & 31; Control: 24 &26). Besides, a negative correlation (0.49) exists between the 2 dependent variables.
Your help is highly appreciated.
Thank you! Dr. Todd Grande, I have a question that what is the allowable value for Mah. distance if I have 11 dependent variables, and what to do if I have outliers?
Hello Dr Todd
May God bless you. Good job fo A great video and explanation.
I have A few question. I follow the video and perform the procedure. Here I got, the data test for outiler, normalities are OK but when looking for the multicolinearity the number is greater than 0.8 and 0.9. What should I? Can I still continue to test the MANOVA?
I see your data is also not normalized for quantity what should I do IF we meet this situation.
Regards
Fahrudin
Is there a non-parametric version of a MANOVA.
I have two lists independant variables (1-3 days, and 6 different concentrations) and 4 dependant variables (which are not normally distributed). When I looked on the internet I could only find kruskall-wallis test but that is not exactly what I'm looking for.
And is there also a test you van do when the dependant variable is ordinal and you have 2 independant variables?
Would we do the same thing for discriminant analysis and logistic regression when checking for outliers?
Hello! A question, where can I find the maximum allowable critical Mahalanobis values? Is there a book or research paper that coul be referenced?
Hi Dr Grande, I have a quick question. The MANOVA is used when you have a single IV and 2 or more DVS. If the DV;s are not obviously related (Let's say, Self-Reported Sleep Quality for one and Sex Satisfaction for 2) is it still possible to run a MANOVA? Thanks
Interesting explanation 👏
What if one instead of quantity you had accuracy on a task, that was measured as a binary of correct or incorrect (1 | 0). Surely a MANOVA would not be appropriate, as looking at means would not produce sensical results, right?
Would you say that to compare that across groups it would be best to split file by condition and run a frequency analysis on the proportion of correct responses? How would you test if those differences were statistically significant?
Can a small sample size effect multivariate normality? There were no univariate or multivariate outliers.
What if we get some multivariate outliers after calculating MD, then what do we do with those data points? Should be retain it or should be remove it?
I have 7 dependent variables and I got a Mahalanobis distance of 31, is this the case for an outlier? If yes, what to do next?
Hallo there, If I report the item wording and compare on the Item level,
do I need to recode my reversed scored items? For running a MANOVA
with negative and positive items do I need to recode my negative items?
Or do I only recode when I want to analyse the factorial level? Thank
you for your help!
Post hoc
Group sizes are equal -> Tukey
Group sizes are unequal -> Scheffe
Thank you Dr. Grande for explaining stats in a very understandable way. Do you have a video with MANOVA using two IVs and two DVs?
Don't you need to conduct a Bartlett's Test of Sphericity in a MANOVA to reject the null hypothesis? Or did I miss it in the video?
Hi Todd,Thank you for this video. Please make a video on how to set up the data on SPSS especially the data from factorial table. I would like to learn how to verify the findings from journal articles if they only give a table of means to work on.Thank you so much!
i tried a similar analysis for the data I have. I would like to ask you one doubt. What is the data type of the column named program that you have assigned? I tried string it didnt work. If I select numeric I cant enter these type of text. Can you please clarify. thanks.
Hi there Todd, thank you for your videos!
I have a question to ask, if my dependent variables for this case are all in Likert Scales.. do i have to compute the variables into one construct and then conduct the assumption testing?
Thank you!
Hi, whom did you cite for cut-of value for correlations? I have a correlations of .826. Most books says the cut off value is .7.
If there some variables have no linear relationship across group then what should we do? I looked up in many books (Stevens, tabachinik you name it!) But nothing is given. Should we go for seperate ANOVA for each variable how can we control the family-wise error rate??
Hi, Dr. Grande. I was just wondering how did you find out that the individual is more effective than the treatment as usual and not vice versa? Thank you!
in mahalanibonis, if we have pre test and post test variable how we determine degree of Freedom in chi square table??
can someone tell me if MANOVA gives you the univariate ANOVAS automatically or you have to run them by yourself ?? Because it gave me some pairwise comparisons for my main effects but not for the interaction. Thank you!
Thank you for the tutorial. Do you know where i can find the Manova critical value table?
Could you tell me what was the research question for this test? Thanks
What are your thoughts about the different theories for sample size? Also, do we need to interpret normality with caution since we are testing univariate normality instead of multivariate normality?
+Ivy Byun I think just like ANOVA the MANOVA is a robust test meaning it is not overly sensitive to normality.
Sooooo useful!! Exactly what I needed. Thanks :)
Sir I have one question
When we conducting experimental study with two level of treatment variable and two dependent variable which is a MANOVA study .
Sir to test the assumptions of MANOVA which scores considered
Pre test score
Post test scores
Gain score
Please sir reply as soon as possible
Thanks sir
Hello Dr Grande
Iam a master student and iam conducting a research on writing development of EFL students across four grade levels
iam rather focusing on syntactic development, so i 'll be using a set of syntactic complexity measures and i 'll evaluate them across proficiency level( grade 1 to 4) . So, basically i've a continous dependent variable ( Syntactic complexity) with 7 levels
and a categorical independent variable ( writing proficiency). Apparentlly i should use one way MANOVA but i faced a problem , though there is a difference between the means of all the syntactic complexity measures examined ( the mean difference is above 3) , post hoc tukey test shows p > 0.5 ( ,99) and some of the mean differences that were significant are acctually in a negative direction, does that have to do with the normality of the distribution of the independent variable or with homogenity of variance -covariance? because according to Box's test the null hypothesis is rejected besides the p in shapiro-wilk normality test is 0.00 . Please if you can explain that to me i'll be so grateful as i'll have to submitt my dissertation in few days and i didn't write the result section yet.
REGARDS
Thank you! This was excellent
Hi, what if I don't meet the assumption of sufficient sample size?
What if I have 14 dependent variables and 1 fixed factor, being a score of 1 or 0. Will that work with MANOVA?
Hi, did you get your answer yet? I have a similar question and if you got the answer for the same, it would be quite helpful to for me to learn about it
How to report Mahalanobis distance results with critical values?
Thank you
Dear Sir, could you give a data file? i try to test the lesson but i don't have suitable data. Thank you, sir!
awesome.. Thanks a lot ..
what to do if the assumptions are violated?
thanks. Very helpful
thank you!
what is the difference between the linearity assumption and the absence of multicolinearity assumption??? aren't we testing the same thing?? Please!
+WOLFPMD3 They are not the same. The correlation doesn't reflect the linearity directly. statpics.blogspot.my/2013/11/correlation-ellipse-matrix.html
the sound is not good I am quite disappointed
marry me