I have a between subject group design where my IV has 4 levels. I am comparing the performance of hypothyroid groups (which are 3: SCH (n=14, OH group n= 15, Euthyroid group n = 12 & HC group = 15) from healthy controls (HC) on different measures of memory. I have multiple measures of verbal memory and working memory. I have been suggested to compute one-way MANOVA. However, MONOVA is a very stringent statistical analysis requiring many assumptions. My data is violating at least three assumptions i.e., normality, univariate outliers & linearity (some of the measures are not correlating with each other). In my research area I have not seen any researcher to compute MANOVA to see the group difference. Most of the researches were conducted using a pre-post treatment design with multiple DVs but none of them have used MANOVA to evaluate statistical difference.
Coming to the question. Will be wise to compute separate ANOVAs for each DVS? If I do compute multiple ANOVAs what measures I should take to account for the type I error while still maintaining a balance for not committing type II error?
Would be great if someone could give me some suggestions.
Thanks
Vatsal
Dear Pandey
From my understanding several ideas have been mixed in your questions, and very important information regarding your groups/random variables were not given. The assumptions of normality, linearity, homogeneity, .. . are also not clear.
Any way, several examples are given in:
. Landau, S. and Everitt, B. S.(2004). A Handbook of Statistical analyses using SPSS.
.Howitt, D. and Cramer, D.(2008). Introduction to SPSS.
Regards,
Zuhair
Thank you Al-Hemyari for your response.
It would be great if you could elaborate on what information you need for a specific answer? Also, the main idea of this question was to whether to go for Multiple ANOVA if assumptions of Manova are not met. The second main idea was to see if there is a way to control type 1 error without commiting type 2 are or false negatives.
MANOVA is usually prescribed for the analysis of the data of such type you have.
As an alternative of MANOVA, you can use ANOVA treating the DVS as one more mode of classification of data.
Thank you Dhritikesh Chakravarty for your reply. If I conduct multiple ANOVA do I have to correct for the alpha level to account for the type I error??
In any test, alpha level technically describes the probability of committing Type I error. It means or describes the probability of taking a wrong decision (more specifically, the probability of rejecting a true null hypothesis). Accordingly, alpha level is selected prior to the test on the basis of how much degree of risk to be happened due to committing Type I error is tolerable or acceptable. Thus the question of correction of alpha level to account for the type I error does not arise if you conduct multiple ANOVA.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA