The aim of my thesis is to see whether anxiety score can predict sleep quality and dream content. So for this, there is one continuous variable that I want to use as a predictor on multiple dependent variables (= anxiety score) to see whether it can predict the outcomes on those outcome variables (PSQI-score, experienced emotions during dreams, etc.). In order to do this, I used the manova function in R:
model1 = manova(cbind(d1,d2)~X
(in this function: d1 = dependent variable 1, d2 = dependent variable 2, X = anxiety score)
Now, when I type summary(model1), I get this output:
> summary (model1)
Df Pillai approx F num Df den Df Pr(>F)
X 1 0.30237 7.585 2 35 0.001834 **
Residuals 36 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
And then, because it is significant, I look at the results per dependent variable by typing summary.aov(model1)
> summary.aov (model1)
Response d1 :
Df Sum Sq Mean Sq F value Pr(>F)
X 1 1791.5 1791.5 6.505 0.01515 *
Residuals 36 9914.2 275.4 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Response d2:
Df Sum Sq Mean Sq F value Pr(>F)
X 1 2164 2164.05 7.3081 0.01041 *
Residuals 36 10660 296.12
Now, I interpreted this as: "so anxiety score is a significant predictor of both d1 and d2, interesting!" But now that I have to report my results, I'm a bit confused on how to do this. Since I want to test linear relationships, shouldn't there be a beta-coefficient to report?
Everything I can find online uses multiple predictors (IV's) with one outcome variable (DV) instead of the other way around, which is what I intend to do...
Any help is very welcome.
I suggest that you consult this text both under MANOVA and Multivariate regression. BTW all of this is much easier if you do multivariate regression(also ) in R. ANOVA was meant for easy calculation and regression for easy understanding. Here are your links: https://1lib.us/book/1047479/fa1894
and: https://www.google.com/search?rlz=1C1CHBF_enUS917US917&ei=B2jqX8zpC8m4tQb24rnoCA&q=multivariate+multiple+regression+in+r&oq=Multivariate+multiple+regression+&gs_lcp=CgZwc3ktYWIQARgAMgUIABDJAzICCAAyAggAMgIIADICCAAyAggAMgIIADICCAAyAggAMgIIADoECAAQRzoECCEQCjoFCAAQzQI6BwghEAoQoAE6BQghEKsCOgUIIRCgAToJCAAQyQMQFhAeOgYIABAWEB5QzmhYmH5g6poBaABwAngAgAGjAYgByQmSAQM2LjaYAQCgAQGqAQdnd3Mtd2l6yAEIwAEB&sclient=psy-ab
Best wishes, David Booth
I don't think that MANOVA is the right choice of statistical test for your data. You write that you have a continuous independent variable (predictor). Are your dependent variables also continuous? And are these dependent variables somehow interrelated or can be tested separately?
Radek Pileček Excellent questions. I apologize for not asking them earlier. As one of my Profs used to say-You know what they say about assume. Thanks, David
Thank you for your answers David Eugene Booth and Radek Pileček . The dependent variables are indeed also continuous. Some of the variables are related in the sense that they were measured using the same scale or methods. Before conducting the MANOVAs, I checked whether the dependent variables tested together in the model were moderately correlated (.20-.60), which is one of the assumptions of a MANOVA. Then I conducted multiple MANOVAs on the variables that correlated to each other. For variables that did not relate to other variables, I used the lm() function.
I also thought that I needed to perform a multivariate regression, but all information that I could find were examples of multiple regressions, in which multiple independent variables were used to predict one dependent variable. Other multivariate examples both use multiple dependent variables and multiple independent variables. Not once have I come across an example using just one independent variable and multiple DV's.
In a talk with another statistics help department, they advised me to perform the MANOVA, as this, according to them, is the same thing as a multivariate regression.
Eva Beunk they are the same but the regression is much easier to understand while the manova is much easier to compute. But if you have a computer ease of computation doesn't matter. Look at the material I sent and make your own decision. But since this is a thesis you want to get it right. Talk to your advisor. That is what S/he is about-helping you. Best wishes , David Booth
No! you don't need a beta (coefficient), reporting the MANOVA (and the individual ANOVAs) is sufficient, the p-value shows the strength of the relationship. Remember, even in the regression analysis, the beta is not as important as the p-value.
Note. should you do regression (separately for each DV) the p-value of the beta will be exact with the one you have in your ANOVA
@Nazir. I am afraid that you are incorrect in several places. Please see the references I suggested above for details on multiariate linear models. Best wishes, David
There is nothing wrong using MANOVA! What is not clear is your data type.
1. If all data variables are of continuous type you are correct.
2. If all dependent are not continuous (one discrete and the other continuous) your analysis is wrong.
3. Hope the two dependent variables are not correlated? If yes then your analysis is wrong.
On the issue of Beta coefficient values I think it is of importance if you are to report the odd ratio of each dependent variables on the independent variable.
Good luck, DIJ.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis