psychology research - what stats to use to analysis my data?

enid

New Member
#1
hello!

I would really like your advice on some research I am doing into teachers' attitudes and self-efficacy toawrds inclusion. Below I have tried to give the information that this forum requests. I hope I have written it clearly!

1. purpose of the study / research question / hypotheses, if possible, provide some background or context for us.
The research questions/ aims of the study is to find out if teachers' attitudes, and self-efficacy towards including children with SEN increase as a result of attending a certain training package.

2. nature of the data - independent and dependent variables
I'm not too sure if I can answer this... I am right in saying that the dependent variables will be teachers' attitudes and teachers' self-efficacy and the independent variable will be the training package? the data will be quantitative.

3. how the variables are being measured
I intend to use an attitude scale before the teachers attend the training course. The training course will last three months. The teachers will complete the same attitude scale after they complete the training course.

the same will happen for self-efficacy. Teachers will complete the self-efficacy scale before the training course commences and then complete the same scale after the training course finishes- 3 months later.

There will also be a comparison group, although this will not be a control group as opportunity sampling will be used to select the particiapnts. This comparison group will only complete the self-efficacy and attitude scales once- there will be no before and after testing here.

- I am trying to find out if there is a significant increase in positive attitudes and self-efficacy if teachers attend the training. However, I am very confused about what stats to use on spss to help me show this. Any help would be very very much appreciated.

Many Thanks

Enid:)
 

CB

Super Moderator
#2
Hi there, welcome.

There will also be a comparison group, although this will not be a control group as opportunity sampling will be used to select the particiapnts. This comparison group will only complete the self-efficacy and attitude scales once- there will be no before and after testing here.
I can understand that randomisation may be difficult for practical reasons, but if the comparison group doesn't also have before and after testing I'm not sure what value there is in making the comparison group measurements at all. What is your reasoning behind not doing before and after measurements for them too? The lack of a "placebo" condition and the lack of randomisation means the design still wouldn't be ideal, but you would at least be making a reasonable attempt at trying to control for things like maturation and history effects if you did have before and after measurements.

- I am trying to find out if there is a significant increase in positive attitudes and self-efficacy if teachers attend the training. However, I am very confused about what stats to use on spss to help me show this. Any help would be very very much appreciated.
The details depend a bit on things like what you end up doing with the control group, but you're probably looking for something like a repeated measures ANOVA. (Analyze > General Linear Model > Repeated Measures). There is a tutorial online here, although the tutorial's list of assumptions for repeated measures ANOVA are incorrect (it assumes normally distributed residuals for small sample inference; the (marginal) distribution of the dependent variable does not necessarily need to be normal).
 

enid

New Member
#3
Re: psychology research - what stats to use to analyse my data?

Hello Cowboybear,

Thank you for your very good reply!

-in response to your first point-

I understand what you mean. I think my reasoning for using a comparison group in the first instance was to establish that there was some consistency between the group who have volunteered to take part in the intervention and the rest of teh population. The reason that I am proposing that this comparison group do not get tested for a second time is for practicality reasons: teachers are so busy and I think realistically they just wont fill in teh same questionnaire a second time- especially if they are not involved in the intervention. I don't want to propose or attempt a piece of research that I will not be able to carry out.

But I see what you mean that this design results in the existence of a comparison group being rather useless.

Would my design be ok if I removed the comparison group all together; if I just look at whether there is a significant increase in positive attitudes and self-efficacy after the test sample receive the intervention?


-in response to your second point...

Thanks for this, it is very clear and very helpful and the website you recommended is also very claer. With what I discussed above, will this analysis stay the same (if I remove the comparison group?)- from what I have read on the recommended website this appears to be the case.

Also you wrote this point "...(it assumes normally distributed residuals for small sample inference; the (marginal) distribution of the dependent variable does not necessarily need to be normal)...."

I'm really sorry ... but I don't understand what this means! would you usually have to test to see if your data is normally distributed?- are you saying that the repeated measures ANOVA doesn't assume the data is noramlly distributed- so I would be ok using this no matter what my data is. Or are you saying that my data will need to be normally distributed in order to use this analysis?

Thanks again for you help, it's fantastic and I very much appreciate it!

Enid
 

CB

Super Moderator
#4
Re: psychology research - what stats to use to analyse my data?

Hi again! Glad you found the reply helpful.

I understand what you mean. I think my reasoning for using a comparison group in the first instance was to establish that there was some consistency between the group who have volunteered to take part in the intervention and the rest of teh population. The reason that I am proposing that this comparison group do not get tested for a second time is for practicality reasons: teachers are so busy and I think realistically they just wont fill in teh same questionnaire a second time- especially if they are not involved in the intervention. I don't want to propose or attempt a piece of research that I will not be able to carry out.

But I see what you mean that this design results in the existence of a comparison group being rather useless.
I can understand that the practicality of two surveys with the comparison group could be an issue. You could maybe alleviate this a bit by making absolutely sure the survey is as short as possible, and also by ensuring an adequate rationale for the study is given to the participants. I'm not sure if you'd necessarily want to tell them too much about hypotheses and so on, but perhaps you could describe the purpose as looking at changes in the DV over time in teachers...?

Would my design be ok if I removed the comparison group all together; if I just look at whether there is a significant increase in positive attitudes and self-efficacy after the test sample receive the intervention?
Well, as you probably know, intervention research without control groups does happen and get reported quite often. The main thing to keep in mind is that without a comparison group you really won't be able to make inferences about the effect of the intervention. You might be able to say that the teachers' attitudes etc changed over the time of the intervention, but you won't be able to make any firm claims that this change was the result of the intervention itself (if such a change happens, it could also be the result of placebo effects, history effects, maturation effects, etc etc). Whether this is acceptable depends a bit on context: Such a study might be quite acceptable for an honours project where the area of study is quite new and exploratory, but less acceptable if the project is quite important and might be used in the future to make decisions about resource allocations to particular interventions and so on.

Thanks for this, it is very clear and very helpful and the website you recommended is also very claer. With what I discussed above, will this analysis stay the same (if I remove the comparison group?)- from what I have read on the recommended website this appears to be the case.
Without the comparison group you will really just be looking at paired samples t-tests for each dependent variable (IV being time).

Also you wrote this point "...(it assumes normally distributed residuals for small sample inference; the (marginal) distribution of the dependent variable does not necessarily need to be normal)...."

I'm really sorry ... but I don't understand what this means! would you usually have to test to see if your data is normally distributed?- are you saying that the repeated measures ANOVA doesn't assume the data is noramlly distributed- so I would be ok using this no matter what my data is. Or are you saying that my data will need to be normally distributed in order to use this analysis?
Your dependent variable data doesn't need to be normally distributed for repeated measures ANOVA. However, your residuals do need to approximate a normal distribution. How important this is depends on your sample size a bit (the bigger the sample, the less of an issue this is).

For a paired samples t-test, the differences between the pre and post scores need to be normally distributed.

Hope that helps :)