Friedmans Test HELP please

[lmay0001]

Not homework (research data)

I am confused over using the Friedman's test. Should the friedman's test be used for the same question ? or should it be used for different questions??

Example:

Q1 1, 3,4,5,5,4,3,2,1,1,1,3,4,5,6,5,6,7,7,7,
Q2 1,3,4,5,3,4,5,5,3,4,1,1,1,3,2,3,1,1,2,3,4
Q3 2,3,4,5,6,76,5,4,4,3,5,6,5,4,6,6,,5,6,7,7
Q4 1,1,1,1,1,1,2,3,4,4,5,1,1,1,1,1,1,1,1,2,4

[victorxstc]

Please give us more detail on your design. By the way, Friedman is for testing the differences of a group for more than two times with itself. So it is not about questions to be compared. But it is about scores to be compares. Moreover, I personally could not at all understand this part:

Example:

Q1 1, 3,4,5,5,4,3,2,1,1,1,3,4,5,6,5,6,7,7,7,
Q2 1,3,4,5,3,4,5,5,3,4,1,1,1,3,2,3,1,1,2,3,4
Q3 2,3,4,5,6,76,5,4,4,3,5,6,5,4,6,6,,5,6,7,7
Q4 1,1,1,1,1,1,2,3,4,4,5,1,1,1,1,1,1,1,1,2,4

What are these numbers? I understand that each line is for one question, but what are the values within each line? Scores of participants? Please give full details.

[lmay0001]

Those are scores given by participants for each question that was stated. A statistitian i talked to told me that Friedman's test can be used. I have to test whether there is a sig.difference between mean scores

[victorxstc]

Then you have 4 questions and 20 participants. So what is repeated here in order to justify Friedman?

Besides, as I stated before

Please give us more detail on your design.

you need too give "details" of your design before we can help you.

[lmay0001]

Attached is the questionnaire.

My population is 218
Sample obtained is 143

Aim: Evaluate the current protocol handbook and investigate the probability of electronic dissemination formats.

the questionnaire aims at collecting data from a population of pharmacists on two dissemination tools

1. protocol handbook
2. Protocol website

I need tests to prove that scores obtained are significant for certain statements over other statements. When i spoke to a doctor in statistics research he told me Friedman's is used as a test to compare the means of scores obtained. through it (via the p-value) u can deduce that the results are not obtained by chance.

[victorxstc]

So you have 143 pharmacists who have answered a number of Likert-scale questions about two protocols. Now you want to check whether some of those questions have better scores over the other ones, and this difference is significant.

In this case, a Friedman will tell you that whether there was any significant difference between the scores of all questions or not. It does not differentiate protocols from each other, so you can't be sure that whether a significant P value calculated using Friedman is pointing to a difference between the scores of the two protocols, or a difference between the scores of the questions within one of the protocols.

For understanding the latter, you have to run a post-hoc.

[lmay0001]

can u suggest what type of post hoc can be carried out between two scores ?? i have no idea believe me

[lmay0001]

would the wilcoxon signed rank test be good?

[victorxstc]

Yes, of course.

[lmay0001]

So just to clarify. If i want to show significance between two protocols example:

Time consuming (non -prescription protocol) Mean score 4.3
Time consuming (prescription protocol) Mean score 2.8

I run the Wilcoxon signed test as the data is paired on time?? Time varies as one uses one or the other protocol. They are related to each other ??

[victorxstc]

No the data are matched on the subject. Each subject has answers for example 20 questions, so his point of view or knowledge etc. will affect the way he answers the questions. So you can match any questions answered by a subject. This is why you can and should run a Friedman first in order to make sure there "is" a difference between the whole data.

I suggest you to sum up the scores of the internet-based and paper-based methods, and run a Wilcoxon between the total scores of each method. Also you can compare the score of any of the questions with the scores of another question (assuming that they are paired on subjects/questionnaires). So you can use this to elucidate the significance of your regions of intereest.

[lmay0001]

Tanks immensely u have saved my dissertation !! haha i got it finally !!!!