# Extreme right skewed data and repeated measures

#### dr.iva

##### New Member
Hi!
I’m writing for the first time and I would be grateful if someone could help me. I am total statistics beginner and additionally I’m a doctor so you can say that statistics is very difficult for me to understand

So. I have a problem with extreme right skewed data that I want to analyze with repeated measures analysis. I understand that I can’t do repeated measures analysis if data aren’t normally distributed, am I right? I’ve done square root and reciprocal transformations but there is still no normal distribution. I wanted to transform data in 3√ but I don’t know how to do this in SPSS (can’t find this option in Compute?!)

Problem is that I really need this repeated measures analysis because this is the only right way (I think!) to analyze my data – there are 2 groups (59 patients and 53 healthy subjects) and we measured one hormone in each patient 13 times (one hour apart).

Any suggestions?

Thank you!

#### dr.iva

##### New Member
Why not use a nonparametric (rank-based) alternative?
Thank you!

I wanted to try parametric test but I will do this. Anyways I don't know how could I conclude anything from some complicated transformation that I tried to do.....

#### alamp

##### New Member
Hi,
I'm Master's student in Biostatistics in Thailand.
Now,I take tutorial class of stat consult and my professor assign me and friends to answer any forum in online forum of 'www.dmbn.net'.I think that your problem have some issue to discuss in our class on monday with my friends and my professor.Then I could invite you to post your question in our forum at www.dmbn.net. At first page you will see the Biostatistics icon, it have a green color,Then click it and in the next page you will find '[biostatistics]' it is the button that you can click it to new post.May be we are able to help you
-Thanks-
Sorry if my english language skill is poor.

Last edited:

#### alamp

##### New Member
Hi,
I'm Master's student in Biostatistics in Thailand.
Now,I take tutorial class of stat consult and my professor assign me and friends to answer any forum in online forum of 'www.dmbn.net'.I think that your problem have some issue to discuss in our class on monday with my friends and my professor.Then I could invite you to post your question in our forum at www.dmbn.net.At first page you will see the Biostatistics icon, it have a green color,Then click it and in the next page you will find '[biostatistics]' it is the button that you can click it to new post.May be we are able to help you
-Thanks-
Sorry if my english language skill is poor.

Last edited:

#### dr.iva

##### New Member
Then I could invite you to post your question in our forum at www.dmbn.net..[/QUOTE] Hi Alamp.... I did it right. Keep me informed ! :wave:

#### dr.iva

##### New Member
new problem !

Why not use a nonparametric (rank-based) alternative?
OK . I preformed Friedman test – I hope that this is right test to do ?!

BUT !! I have another problem.

As I previously said, I have 2 groups (patients and controls) and each participant has 13 repeated measures (one hormone measured from the blood sample form 09h – 22h). Problem is that I KNOW (every doctor does ! ) that this hormone level changes during this time in both groups (it’s normal!) but I assume that patients have overall lower level of this hormone in comparison with controls.

So. I can do Friedman test in 3 ways:
1. both groups together
2. Only in patients
3. Only in controls

No matter how I do Friedman, I have significant result which tells me that there is a difference in this 13 measurements. I suppose that normal hormone “variability “ during the time in 2 and 3 option (only patients or only controls) is responsible for test significance.

But what about the 1.option (both groups together)? How can I exclude “normal variability” of this hormone level during the time?

Sorry for such long message, I really need to do this right !
Thank you !

#### alamp

##### New Member
Hi dr.iva
I see your posted in my forum.
I will coppy some text in the above box that your posted to my forum.
Because,It's append information,that useful for discuss.
-Thanks-

Last edited: