Can anyone direct me towards someone or somewhere when I can get simple advice on the right tests to use in SPSS?

coyote

The 'right tests'? A worrying statement.

Many Universities (well mine does) has a section that provides stats advice to grad students.

Unfortunately mine doesn't. And my supervisor isn't interested because this is for some research I did before starting my PhD

You up early for a round of golf? Only golfers get up this early in the morning

coyote

If you don't have a distinct stats service, is there not someone in your department or in your uni who knows something about stats who can help?

I'm sure there is someone at your uinversity you could chat to - but otherwise there are tons of SPSS manuals around which have chapters on choosing the most appropriate tests.

Almost any basic statistics book has so called "decision chart" or "decision tree". There are diferent questions about your data and answering to them you are lead to the right statistical test.

I am a statistician and it is really hard to advise how to best analyse data wihtout actually seeing it. Your best bet would be to contact your uni's maths/stats department and see if there is anyone who is willing to advise you how to proceed. They will consult with you on how best to analyse the data but don't expect them to do it for you. Most stats people are really nice and get excited easily by new data, so am sure they will help and advise you.

Hope this helps.

Hi and thank you for your replies. I have taken advise with two stats people and have been told I need to apply the Krusal thingy test. I am looking to see if there is any significant difference in the amount of stress - measured on a scale of 1-4 - experienced by students from five different Unis. I have averaged all their scores and want to compare these. If there is no statistical diff I can then discuss them as a generalisable whole.

But - I cannot seem to make that test work right. And this is where I now need help. And everyone is too busy and I want to do this for a conference at the end of next week, smile.

So I have 30 factors B1-B10 C1-C10 D1-D10 and five campuses 1-5. What do I put in each box etc?

Linda

OK, so a Kruskal-Wallis Test looks to see if there are any differences in medians stress level between the groups, i.e. universities. The median is just an average like the mean.

A really great paper on the Kruskal-Wallis test is:

Benwick,V, Cheek,L & Ball,J (2004) Statistics Review 10: Further Non-Parametric methods. Critical Care. Vol 8, No.3 pg 196-199.

The stress levels are your dependent variable and you want to put universities as a factor. The liscence to my version of SPSS has expired so I can't look to see what options there are for a K-W test. Is this any help?

I found the book by Julie Pallant to be very useful for this sort of thing - she gives you step by step what you should put in etc - not too expensive and well worth investing in for any aspect of SPSS

Sorry to keep posting, I was right about most stats people being willing to help

http://www.utexas.edu/courses/schwab/sw388r6/SolvingProblems/KruskalWallisTest.ppt

Gives a step by step guide to doing this in SPSS with examples ans screenshots.

I can't say thank you Silly Billy - it sounds silly. I am Linda.

The test says test variable and then group variable and then you have to define the range. I have tried to put the title of the column B1 in the test variable and the group variable in as campus and the other way round but I get something about empty cells for one and nonsence answers for the other.

Sorry to be a pain. Was meant to have a stats to help me in this project but as this one is not a part of my PhD but something i wanted to do for work suddenly they are too busy.

Linda

Your data needs to be in this form:

Stress Levels Campus
1 1
2 1
1 2
2 2
. .
. .

So the 'Campus' variable needs to be numeric with 1 indicating the first uni, 2 the second and so on. The 'test variable' is stress level column and the 'group variable' is your 'campus' variable.

You then need to click on the 'define range' button and enter numbers 1 and 5 (because you have 5 universities in your campus variable). Make sure the test type is set to Kruskal-Wallis H and press OK and you should be ok.

Billy

Sorry, I don't think that i have made the form of the data clear. The first column needs to contain all of your stress levels. The second column (campus) need to indicate which university the student is from i.e. 1, 2, 3, 4, 5. The stress levels need to be the raw stress level for each student and not an average.

Billy I could kiss you!!! Thank you Thank you Thank you.

Of course some of them are significant but not many so I can deal with all these separately.

Where do you do your stats? I am in Scotland.

Linda