Which stats test to use

 Posts: 17
 Joined: Tue May 15, 2007 6:36 pm
Which stats test to use
Dear Stats Friends
I am a Clinical Psychology Trainee carrying out a small (5000 words) service improvement project based on where I am on placement. The title of the project is:
“Quantifying and comparing uptake of CBT therapy for clients with psychosis in four complex needs teams”.
The aim of the project is to see what percentage of clients are receiving 16 sessions or more of CBT for psychosis as per National Institute for Clinical Excellence (NICE) guidelines.
After analysing the data I created 6 categories of therapeutic delivery:
1 received 16 or more sessions
2 received < 16 sessions
3 Therapy Ongoing
4 Intervention declined by client or service
5 No intervention offered
6 Family Therapy
The descriptive analysis tables shows the percentage of each team’s clients who have received each category of therapy per team. So for example in team 1, 60% of clients received CBT>16 sessions, team 2: 40%, team 3: 50%, team 4: 10%.
So the team is the independent variable and % of uptake of CBT > 16 is dependent variable. I now need a statistical test to tell me if there are significant differences between the scores for the four teams and if so where they are. I assume I will run the statistical test for each of the 6 categories.
Initially at the proposal stage, my supervisor suggested Chi squared as I don’t need to know the degree of significance but given that there are 4 teams is this possible to do? From my reading it seems one way ANOVA plus post hoc tests (not sure which one) is what is required. Any advice appreciated!
I am a Clinical Psychology Trainee carrying out a small (5000 words) service improvement project based on where I am on placement. The title of the project is:
“Quantifying and comparing uptake of CBT therapy for clients with psychosis in four complex needs teams”.
The aim of the project is to see what percentage of clients are receiving 16 sessions or more of CBT for psychosis as per National Institute for Clinical Excellence (NICE) guidelines.
After analysing the data I created 6 categories of therapeutic delivery:
1 received 16 or more sessions
2 received < 16 sessions
3 Therapy Ongoing
4 Intervention declined by client or service
5 No intervention offered
6 Family Therapy
The descriptive analysis tables shows the percentage of each team’s clients who have received each category of therapy per team. So for example in team 1, 60% of clients received CBT>16 sessions, team 2: 40%, team 3: 50%, team 4: 10%.
So the team is the independent variable and % of uptake of CBT > 16 is dependent variable. I now need a statistical test to tell me if there are significant differences between the scores for the four teams and if so where they are. I assume I will run the statistical test for each of the 6 categories.
Initially at the proposal stage, my supervisor suggested Chi squared as I don’t need to know the degree of significance but given that there are 4 teams is this possible to do? From my reading it seems one way ANOVA plus post hoc tests (not sure which one) is what is required. Any advice appreciated!
Re: Which stats test to use
My first thoughts were chi square, I don't think you can do ANOVA's with percentages as it looks at means but i may be corrected by someone with more stats knowledge than me!

 Posts: 17
 Joined: Tue May 15, 2007 6:36 pm
Re: Which stats test to use
Hi Thanks for your reply
When I looked at Chi squared it requires you to enter observed results and then expected results and I have no idea what they would be? It's not like throwing a dice when you can calculate the probability. Some categories are much smaller than others e.g. Therapy ongoing vs no therapy offered. I thought ANOVA because it is basically the equivalent of multiple T tests but you may be right about the percentages. It makes no sense to use raw figures because some teams have more clients than others so raw figures aren't comparable...
When I looked at Chi squared it requires you to enter observed results and then expected results and I have no idea what they would be? It's not like throwing a dice when you can calculate the probability. Some categories are much smaller than others e.g. Therapy ongoing vs no therapy offered. I thought ANOVA because it is basically the equivalent of multiple T tests but you may be right about the percentages. It makes no sense to use raw figures because some teams have more clients than others so raw figures aren't comparable...
Re: Which stats test to use
Why don't you turn it into a binary logistic regression.
So have all the patients, and give them 1 or 0's representing CBT or not CBT. (dependent variable)
Then 1 to 4 to put them into categories.
Then run a binary logistic regression which will show you whether the group membership predicts uptake of CBT.
So have all the patients, and give them 1 or 0's representing CBT or not CBT. (dependent variable)
Then 1 to 4 to put them into categories.
Then run a binary logistic regression which will show you whether the group membership predicts uptake of CBT.

 Posts: 93
 Joined: Mon Mar 10, 2014 7:46 pm
Re: Which stats test to use
The expected results in the chi square would be the average uptake across all teams, I think.
so if team A was 40%, team B 60%, and team C 80%, the expected would be 60%.
But this is just off the top of my head, so may be way off!
so if team A was 40%, team B 60%, and team C 80%, the expected would be 60%.
But this is just off the top of my head, so may be way off!

 Posts: 17
 Joined: Tue May 15, 2007 6:36 pm
Re: Which stats test to use
Aha sparklehead. You are a genius!
My uncertaintly was exactly over what the expected values should be. In the examples I had looked at, either the expected values were a probability calculation e.g. comparing results of 100 rolls of a dice with what probability would predict. Or based on existing information, so in the Chi squared test looking at relationship between high and low academic achievement and low, middle, high self esteem, the expected values of self esteem probability are based on research which shows 20% of population are low, 30% are high and 30% are high. With regard to my own data, I didn't know what expected results should be. The only thing I could think of was to divide the overall percentages by the 6 categories so expected result for each category would be 16.66%. My concern with this was that it did not reflect the reality which is that some categories could be expected to be bigger than others.
Is there a link you can point me to about expected results being an average of observed results?
My uncertaintly was exactly over what the expected values should be. In the examples I had looked at, either the expected values were a probability calculation e.g. comparing results of 100 rolls of a dice with what probability would predict. Or based on existing information, so in the Chi squared test looking at relationship between high and low academic achievement and low, middle, high self esteem, the expected values of self esteem probability are based on research which shows 20% of population are low, 30% are high and 30% are high. With regard to my own data, I didn't know what expected results should be. The only thing I could think of was to divide the overall percentages by the 6 categories so expected result for each category would be 16.66%. My concern with this was that it did not reflect the reality which is that some categories could be expected to be bigger than others.
Is there a link you can point me to about expected results being an average of observed results?
Re: Which stats test to use
Chi square is with counts  the number in each group (not the percentage) and this is compared with the expected proportion. The expected figures could be based on the mean uptake of CBT across all the teams. It absolutely does make sense to use the counts and not the percentages, as that means the counts are being compared to the expected proportion (which will be a higher count in larger teams).
But really I think you need to divide it up into two analyses. The values that are important are the under and over 16 sessions counts or offered a therapy that wasn't CBT, so that first chi square should only include completed treatments. You can then do a separate chi square on whether more treatments are declined in any particular teams. And if you wanted you could also compare CBT vs non CBT in current and completed treatments.
But really I think you need to divide it up into two analyses. The values that are important are the under and over 16 sessions counts or offered a therapy that wasn't CBT, so that first chi square should only include completed treatments. You can then do a separate chi square on whether more treatments are declined in any particular teams. And if you wanted you could also compare CBT vs non CBT in current and completed treatments.

 Posts: 93
 Joined: Mon Mar 10, 2014 7:46 pm
Re: Which stats test to use
Miriam is right  it is about counts actually  but that's easy to suss out too. I always tend to google statistics as I'm doing them. http://math.hws.edu/javamath/ryan/ChiSquare.html might be of use. I think it's pretty straightforward (maybe)  yay! good luck
Who is online
Users browsing this forum: No registered users and 1 guest