## About Easystats

Easystats is aimed at the novice health services researcher who usually has little or no knowledge of statistics and how they might be used in their research project.

The best way to use this blog is to view the posts and categories to see if a question or concept has already been questioned or explained. If you can’t find what you are looking for then post a question. Alternatively, if your question is partly explained, then post a comment asking for more information.

Never think a question is too simple to ask, but if you don’t want your question to be made public, then email me directly using the email address below. Remember, this is a new blog so much work still needs to be done and is ongoing. This blog will only work if people ask questions.

Easystats is maintained by:

Dr. Jim Turner

Senior Research Fellow

Clinical Audit, Research & Development

Betsi Cadwaladr University Health Board

Phone: *01978-727502*

E-mail: jimzstats@gmail.com

Please post a question or comment in the box below ðŸ™‚

I have got2 groups, 7 variables and four levels of measurement. I am using MANOVA to look for significant differences between the two groups. Can I use the Univariate tests output in the MANOVA output to identify where the differences are. Or do I need to run ANOVA with each variable.

Does this make sense?

Colin

The univariate tests should be adequate

Hey, thanks for the site.

Is there an easy way to test what distribution family my data follows?

Thanks

It depends initially on what type of data you have. If it is continuous then you would firstly plot the data points using either/both a histogram or boxplot to get an idea of the shape of the distribution. You would be looking to see if it follows a normal distribution; you can test for normality by using tests such as the Kolmogorov-Smirnov test, the Anderson-Darling test, or the Shapiro-Wilk test.

If your data is nominal or ordinal, then your data is unlikely to follow a normal distribution, which relies on being able to calculate a mean and standard deviation; having said that, given a sufficient number of degrees of freedom, even this kind of data can approximate a normal distribution.

Again, you would plot the data to get some idea of its shape, to be followed by an appropriate test. The chi-square goodness-of-fit test may be a good start as the test requires that the data be divided into categories. This is appropriate with discrete data, which can take on only a small number of values.