How to enter data for lilliefors test

Hi,

I am looking for a normality distribution test for small samples. Research has lead me to believe that the lillietest function will do the job, however I am unsure of how to out the data in? Effectively I have collected two EMG signals under two different conditions. Matlab help file indicates the input format is [h,p] = lillietest(x). Based on this should I perform two tests for normality each with two signals? i.e. x = condition one y = condition two?

Or should I put all data together and run the test for normality only once?

I appreciate any help or advice that can be offered. Thanks Dan

Answer 1

[Wayne King]

Hi Daniel, Yes, lillietest() takes a vector input, so you will need to input the signals one at a time.

[Daniel Robbins]

Hi Wayne thanks for answering, just to confirm though, do I need to run a test for each individual signal i.e four tests; or run a test for each condition i.e two tests.

Thanks again

[Wayne King]

Hi Daniel, from what little I know of your experiment, it appears to me that you need to run it for all four. I mean what would you conclude if you the two time series for the same condition gave different answers?

Also, you should some correction (Bonferroni for example) for your multiple tests.

[Daniel Robbins]

Hi Wayne,

The overall aim of using this function is to determine if the data is normally distributed in order to ascertain whether to use parametric or non parametric statistical tests.

The EMG signals were obtained both in the presence and absence of vibration. In reality they are also groups of mean data selected at a specific point in time i.e. at a certain point in a movement. For example I have data from two muscles not exposed to vibration, then additional data from the same muscles exposed to vibration.

Therefore I am assuming if I run lilletest on both conditions i.e. vibration and non vibration I can work out if either has a non normal distribution. If so then I will apply non parametric testing.

Based on this additional info would you still suggest I run four tests?

Hopefully that makes some sort of sense.

Thanks again for you input, its much appreciated.

dan