Im doing a big project and its been a while since I've taken a stats class. This is all hypothetical since I just have to create a program but dont have to actually implement it. I have created a questionnaire which I will distribute to students before and after the implementation of my program. I need to run an analysis on the before and after results of this data once they have completed the intervention program. Im also using another school as a control group so they will be getting no intervention so I also want to run an analysis on the clery reports of their school between now and in 2 years and then also on my schools offending rates from now and in 2 years. What kind of statistical analysis would I use and why? Regression? Correlation? Pearsons r? etc... Any help is appreciated and please let me know if anything needs clarifying

Run an independent group t-test.

Also try using program called SPSS which does all the calculations for you and lets you analyse the results.

Well I dont need to actually run any tests. Im creating a proposal for a alcohol prevention program that I came up with and there is a part where it asks what type of statistical analysis would I run and why.

So if independent group t-test is your choice what is the reason why?

Other opinions are greatly appreciated too!

No, your samples are not independent, so a t test is not appropriate, because school 1 may differ from school 2 in initial conditions. You need to account for pre- and post effects as well as effects between schools.

More later when I have a few minutes for a proper answer.

Ok great. If you need clarification on anything let me know. Basically gonna be analyzing the before and after survey data and then also the clery reports between the 2 schools in 2 years

So you're surveying the entire population of a school? Just those students who are likely to become criminals and who might be helped by the intervention? It's not clear how the survey, the intervention, and the Clery reports tie together. If you articulate the question/hypothesis as clearly as possible it'll be much easier to propose the correct analysis.

Understandable. Basically we made up an intervention program thats going to be run through the school and help enforced by the court system in our county. Any student who is in court for his 3rd offense will be processed through the court system and then placed into our program. It is here where they will have to take an alcohol questionnaire to see what their habits are. So just the students in the program will be taking the questionnaire. After they complete the program, they will be taking the questionnaire about 6 months or so afterwards to see if their habits changed. I want to be able to analyze the changes in this data.

Also we want to compare the # of alcohol violations between two alike schools, one that receives the intervention and one that doesnt. We have the clery report numbers now for both schools now and in 2 years we would have the new numbers. I wanna be able to analyze this data to see if the amount of violations went down due to our program. This may just be looking at the clery reports needing no analysis but Im not sure.

Hope this clears things up! Thanks for your help!

OK. So, the Clery data...probably no analysis is fine, since you're only dealing with summary numbers.

The questionnaire...you have a pre- and post, and the way you describe it in OP, you have a control group. In your last post, you describe not having a control group. Obviously it's better to have a control group. What you use to populate the control group (same school or different school) depends on the specific question (hypothesis) posed.

In either case, there are two contrasts: treatment effects and time effects. And there are either one or two response variables: attitudes(?)...some quantity derived from the survey, and actual violations, from the Clery report. It's likely that your factors are correlated as well as your response variables.

So, your approach most likely will be a repeated measures anova done multivariately (i.e. there will be both within- and between-subject tests done). It's not too difficult, but it is implemented differently in different packages. In SPSS, it's implemented through GLM (General, not generalized, linear models)...multivariate. Other packages will have similar implementations.

If you need conceptual explanations, search "profile analysis" or read the appropriate section in Tabachnik and Fidell, which your library will have. It's not a technical book at all.

Thanks alot for the response. Everything is hypothetical as if I were going to actually implement the program so just knowing what type of analysis to run and how I was going to run it is plenty good enough

the explanation of what to do all seems fine. but i'm still really confused on this whole concept of people still using SPSS

SPSS (or SAS) is the de facto standard for many fields. R may be zomg the greatest, but the learning curve isn't worth the effort when processes are already in place and budgets (=available manpower) are incredibly lean.

R was a lot easier for me than SAS was

spss is in the process of dying. if you don't want to use R, stata is the preferred package in most social sciences.

also, since the advent of rstudio, getting going in R is massively easier

^^^so much fail in those statements. Guess I'd better unsubscribe from my spss listserv and tell the 3 universities I've worked for to STOP THE MADNESS. And I'll have my wife tell the analysts at T Rowe Price to switch too since they're obviously on a fool's errand.

if they have 10k+ lying in between the couch cushions to buy SPSS, or PASW, or whatever it is being called now, god bless. for my money, stata provides a more flexible and powerful environment and is more stable. The only reason I'd consider SPSS or SAS is if I was working with 1GB+ datasets, and even then I'd probably use R

I'm not disagreeing with anything you're saying. R is a better, cheaper alternative with a helluva learning curve for 98.5% of the population. (and probably 20% of that 1.5% is here on 2p2.)

In grad school, exactly 2 of the 30ish students in my group learned R, because we simply didn't have the time. I'm not arguing what's better; I'm arguing what the standard is. And it's not R, it's not stata. It's SPSS and/or SAS. If SPSS weren't a money maker, why did IBM buy the company 3 years ago? Do you really think they thought "Oooh, here's an idea. Let's buy a company that's going to be doomed in 10 years and totally unprofitable when it falls to the freeware that is clearly superior?"

Also, thanks for derailing this thread on something that has little to do with OP. I'm sure he's stoked to spend 6 months learning a package he'll never use again in his life, when he can point-and-click his way to semi-right answers that might provide just enough understanding of the stats involved to get him where he needs to be.