SPSS/Statistics Help!

c104

Im currently doing some research, Ive collected my data and now am at a loss of how to analyze it!
I have very limited knowledge of SPSS...
I would really appreciate if anyone could offer me some advice, I've tried reading Andy fields book, but I'm still confused.

My independent variable has 2 possibilities (in or out)
My dependent variables include:
1. a length of time in months, ranging from 1 month to 10 months...
2. where the person was 1st seen, (there are 2 possibilities here)
3. Another independent variable is how they were referred (there are 3 possibilities here)

What type of statistical tests should I use??
Any help would be much appreciated!!

2Scoops

What exactly do you want to do? It's not clear if you want 2 separate analyses on your 2 dependent variables or construct a model with all variables.

c104

Id like to do separate analyses on each of the dependent variables.
But I just don't know which tests I would do on them. I presume they would both need a different type of statistical test?

2Scoops

Sorry, I still don't know what you want to do. I'm not even sure if your 2 dependent variables are actually dependent. One of them looks like it might be another independent. Perhaps it would be simpler you if you just stated in plain English exactly what you want to know e.g. "Are waiting times shorter from different kinds of referrals?" etc.

Come up with a simple question like that for each analysis you want to do and it will be much easier to give advice.

c104

Hey 2scoops.
Sorry I realize that was a little confusing.

So I want to compare 2 referral drops and the time to being seen by the 2 referral groups.
i.e: Are group 1 seen quicker than group 2 because they're referred via a different route.

So i will have group 1 all referred at different times, and it will be between 1 and ten months until they're seen, and I want to compare these to group 2 who have all been referred via a different pathway, and they will all also been referred at different times and will also take a number of months to be seen.
I want to see which group are seen quicker and what is the difference in time to being seen.

Does this make sense?
Thanks for all your help, i really appreciate it!!

2Scoops

c104 wrote: »

I want to see which group are seen quicker and what is the difference in time to being seen.

This is a very straightforward one: your dependent variable here is time, your independent variable is referral route. An independent samples t-test is easy, suitable and will tell you if there is a significant difference between the means. To keep everything above board, you should also establish that the data have a normal distribution and that there is homogeneity of variance between the 2 groups. However, the t-test is particularly robust to violations of these assumptions, especially when the sample sizes in both groups are equal.

These are all easy to do in SPSS but I can show you how if you're not able to find out on your own.

Your first post suggests there's a bit more to it than just this comparison, however. You will need a different (but very similar test) if you are comparing 3 groups and will need a non-parametric test if your dependent variable is one of your nominal variables.

c104

Thanks so much for this.
Its beginning to make sense now... I think i should be able to figure it out on spas as the book i have shows me step by step
However my sample sizes are not equal in the 2 groups. Does this change my test?

2Scoops

c104 wrote: »

However my sample sizes are not equal in the 2 groups. Does this change my test?

No, but it means you're treading on slightly thinner ice when it comes to violating assumptions. It's not something I'd worry about until you know what you're dealing with, results-wise.

opinion guy

c104 wrote: »

Thanks so much for this.
Its beginning to make sense now... I think i should be able to figure it out on spas as the book i have shows me step by step
However my sample sizes are not equal in the 2 groups. Does this change my test?

What are your sample sizes ?

c104

They're pretty small unfortunately
Theres 14 in 1 group and 10 in the 2nd

Ive also plotted my data and they groups are both skewed (not a bell curve)
My limited understanding means that this is not parametric...therefore i can't use a t test ? Is this correct? What would be my next option?

Sorry if again I've stopped making sense....

opinion guy

No no, you are making sense (to me anyhow - its possible we are sharing a delusion however :P)

Yes you probably shouldn't use a parametric test like t-test - both because the numbers are quite low, and because you say the data is skewed. Skewed data means its not normally distributed. There are things you can do to normalise it - but given your low numbers I think so called non-parametric tests are more appropriate.

The word parametric by the way, refers to the assumptions underlying a test (or parameters required). It doens't refer to the shape of the distribution. Bascially your distribution is skewed....and therefore not normally distributed.

That leaves you with non-parametric tests. These are less powerful but require less assumptions (such as the assumption of normality required by t-tests).

I'd probably use the Mann-Whitney U test for the data you are describing.
http://en.wikipedia.org/wiki/Mann%E2%80%93Whitney_U

c104

Brilliant thanks a million.
Ill give the mann whitney u a go!
Ill post again if I'm having trouble.
Thanks!

evolving_doors

Sorry to butt in on this thread but i also have a query regarding SPSS, I thought about starting a new thread but figured it would be more useful to keep it together here..

So basically I have Likert scales that I want to visualise on some chart. (About 7 Statements... Strongly Agree,Agree,Disagree,Strongly Disagree...
The sample size is fairly low (n=33).
Hopefully be able to read the values of each chart too.

Thanks, p.s. I tried googling it a good bit but can;t get anything. Also have Paul Connolly's book but it;s a bit 'highfalutin' for me at the mo. I just want a nice picture!!!!

2Scoops

You want to ideas on how to present the data or just how to use SPSS? For the first question, I would say it depends on the complexity of your analysis. A simple bar graph would be suitable if there is just one group being surveyed, and if more subgroups are being compared maybe try something like this staggered Likart visualization.

evolving_doors

Cheers yes I want to do a list of the questions with a horizontal bar beside each one , each bar contains four colours representing the size/portion of the response. With the ability to read the values exactly.

the link you gave was fairly interesting (although not specifically for spss I think) as it was able to assign the responses to particular groups of people..i.e. The people between 20 and 25 strongly disagreed with the following statement the most compared to any other age group. This would be my next level of analysis, first paragraph is my first priority.

Basically i need a walk through the steps in spss... Ive done pie charts histograms cross tabulations etc. but can't for the life of me get this working,
Thanks a million..

evolving_doors

Armelodie wrote: »

Cheers yes I want to do a list of the questions with a horizontal bar beside each one , each bar contains four colours representing the size/portion of the response. With the ability to read the values exactly.

the link you gave was fairly interesting (although not specifically for spss I think) as it was able to assign the responses to particular groups of people..i.e. The people between 20 and 25 strongly disagreed with the following statement the most compared to any other age group. This would be my next level of analysis, first paragraph is my first priority.

Basically i need a walk through the steps in spss... Ive done pie charts histograms cross tabulations etc. but can't for the life of me get this working,
Thanks a million..

Anyone??

Ostrom

If I'm understanding you, you need a stacked bar chart which can be produced from the chart builder menu (drag the type of chart into the white area, set the bars to proportions - i can't remember the text of the exact option in spss - then drag your variables from the left side variable list). Alternatively, throw the proportions from the crosstab into excel and you can do it a little quicker. Also (this probably won't help, but in case you plan on doing more) there are many google-able commands for stata to get these kinds of chart; they look a bit nicer and are more easily customizable. There is a dedicated book on stata graphics also which contains a section on likert visuals, although if this is a once-off, it is probably not worth the trouble.