# Desperate call for statistics help

#### GeorgiaC

##### New Member
Hello, I've just finished my placement year in my undergraduate degree but I'm really stuck with which statistical tests I need, and how to input the data into SPSS. Any help would really mean the world.

Introduction to study: Working within a community mental health team. Researching whether there is a difference between the recommendations of therapies given by clinical psychologists and nurses.

Psychologists and nurses rated the same clients as to whether certain therapies would be suitable eg:

Psychologist: Client 1: Therapy 1 yes Therapy 2 no Therapy 3 yes
Nurse: Client 1: Therapy 1 yes Therapy 2 no Therapy 3 no

IV: which group of practitioners the ratings were given by.
DV: which therapies were recommended. (I think?)
Example of raw data is attached.

Please ignore the individual or group therapy part, these were then expanded out to their own categories eg. Therapy re trauma individual, therapy re trauma group etc.
What I am aiming to find out if whether there is a significant difference between the two groups of practitioners for the same clients. Any help on this really would be so appreciated. I’m so stuck.
Thank you so much.

#### noetsi

##### Fortran must die
I am confused by how one of the variables is coded. You mention "Psychologists and nurses rated the same clients as to whether certain therapies would be suitable" which suggest more than one therapy is possible, then in the data you either have therapy or no therapy. Is the therapy variable just therapy/not therapy, or is this variable coded with a number of different values (3 or more)?

If you have two variables and you are trying to determine if one of the variable level depends on what the other variable level is, than the most common test is probably a chi square test of independence. If you run this in SPSS you will get a p value. If this is below some value (normally .05) you conclude that there is a statistically difference between the levels of one of the variables in terms of the second variable.

http://en.wikipedia.org/wiki/Pearson's_chi-squared_test

Go to the test of indepence section.

#### GeorgiaC

##### New Member
Hi,
It is possible for more than one therapy to be chosen yes. I'm trying to look at whether in general the 2 groups of people agree. The therapy variable is essentially: is this therapy needed: yes/no and then there are around 20 therapies for which this question is answered.
I'm sorry if I worded it badly before. Is this any clearer?
Thank you so much for your time.

#### noetsi

##### Fortran must die
Are you trying to determine if 1) the different group of respondents vary on whether therapy is needed or not or 2) if they disagree on which therapy to assign when they feel a therapy is needed or 3) both.

#### noetsi

##### Fortran must die
Both can be done per my previous post with a Chi Square Test of Indpendence. This is a standard test that all software supports and well known. The limit is that if you have a lot of therapies, especially with few cases you may encounter difficulties. In this case the solution is often to collapse the therapies into blocks with each block a group of logically associated levels. For example you might collapse age (each year being a separate level) into 1-17, 18-34 etc.

I can not tell you how you would collapse the therapies into blocks, but there should be literature or perhaps your test that does. And of course this might not be an issue depending on how many therapies you have. I suggest looking at the link I posted as a start point.

#### GeorgiaC

##### New Member
Thank you so much. Will do.