A little help with choosing tests?

#1
Hello,

I am new here. A little background, I am a Masters of Forensic Science student with a little statistics knowledge. I am currently working on my thesis and am not sure which tests to use to test my data.

In my first set of data I have two time frames with data which contains 8 columns of data. I have run a 2x8 chi squared that shows their is significance at p<.001. I am specifically interested in one of the outcomes and would like to test if it has changed significantly in regards to the other 7 columns and the total of all columns. Not sure how to proceed.

Secondly I have several years of data that are each associated with the 8 columns mentioned previously, plus data that shows 5 possible outcomes from the 8 columns. I would like to see if there is any correlation between one column and the 5 outcomes.

I hope this makes sense. Can anyone suggest what tests would be adequate? I have a general idea, but I am not positive.

Thanks
 

Karabiner

TS Contributor
#2
Could you perhaps describe what this all is about? i.e. what
was the topic/research question, how were the data collected,
what are the variables measured, how large was the sample size?

With kind regards

K.
 
#3
Sorry, let me be more specific.
The first set of data I have consists of two different date ranges (recent and older), and each of these has data for types of force used by police. Of specific interest is the data on Taser. I would like to know if it use has significantly increased over time. I would also like to know if others have significantly decreased over time.

I am thinking a chi-squared 2x2 of each type against the total for the date ranges?

For these date ranges I also have injury outcomes for each year divided into 5 types of injuries. I have already got correlation coefficients showing that two types of force correlate strongly with certain types of injuries. I just need to get the p value for those.

Hope that helps.

Thanks
 
#4
Ok, let me take this one question at a time.

If a have a list of data for 10 years. How would I test if the number of one outcome has significantly increased over the years when compared to the total outcomes for each year?

Can anyone help please?

I have attached part of my data if it helps.

Thanks
 
#5
If you have data for 10 years on some incident (such as use of Taser) then the data look something like this:

2000 100
2001 110
2002 105
....

right?

In this case, you could do a regression with "Number of taser uses" (or whatever it is) as the dependent variable and "year" as the independent variable. If the variable is "number of incidents" then it is technically a count and you should do Poisson or negative binomial regression, but if the counts are large, then it won't make much difference.

However, if your DV is a proportion then you probably want logistic regression.
 
#6
Thanks for your reply. Yes, that is how the data is structured. I also have other columns of other types of use of force. So is there also a way to determine if the taser use has gone up when controlling for the total number of uses of force each year.

I know that taser use has gone up, but so has the total use of force. Would I be correct if I analyzed for a significant linear relationship, and if it is linear with the total could I state that taser use hasn't increased significantly in relation to the total use of force?

Also, the data are counts.

Thanks
 
#8
Hey everyone,

which test do I use to measure if my responses were significantly better than guessing?

My supervisor said that I need to check whether participants did overall better than chance.

I tried to google a bit but only found cohen's kappa and I'm not sure if I can use that?

Thanks a lot!