Guns and people shot: two pieces of data. How to handle them?

parc

Ok I've got two variables; amount of guns owned and people shot (in selected American states)

What statistical methods should I use to explore these variables. So fair I've already used:

- Scatter plot (as one goes up the other goes up)
- Box plot to identify any outliers

Anymore that I could use? Is there a statistical test that I can use to test the reliability of the relationship?

You don't have to explain it, just tell me what I could use and then I'll go off and learn how to apply it

Thanks!

bnt

If you do the scatter plot, you can calculate a regression line to go with it, and a coefficient of conformity (R²) which tells you how well the line fits the data.

If you do the plot in Excel, you plot the regression line by saying "Add Trendline", with the option to print the R² value too.

parc

bnt wrote: »

If you do the scatter plot, you can calculate a regression line to go with it, and a coefficient of conformity (R²) which tells you how well the line fits the data.

If you do the plot in Excel, you plot the regression line by saying "Add Trendline", with the option to print the R² value too.

Ah thanks that's really helpful. I was going to do it in SPSS but it's good to know how to do it in excel.

MathsManiac

By the way, are these absolute numbers of guns owned and people shot, or are they "per unit population"? If it's the former case, you could find that there's a correlation only because both are related to the population of the state, and this would hamper the interpretation. (In this case the population of the state would be a confounding variable.)

parc

Hi there, is not per head. Is there anyway round this. Some way to weight it differently

Also what is a decent R2 value to show that it fits to the line. I heard around .5 is good.

.7 is great

Michael Collins

MathsManiac wrote: »

By the way, are these absolute numbers of guns owned and people shot, or are they "per unit population"? If it's the former case, you could find that there's a correlation only because both are related to the population of the state, and this would hamper the interpretation. (In this case the population of the state would be a confounding variable.)

The classic example here being that on days where ice cream consumption increases, drowings also increase. Therefore ice cream causes drowning*.

*Or so it might appear. Of course the weather is the confounding variable here, and in this case the premise is noticablably absurd but as Mathsmaniac has pointed out, often it's not so obvious that a third variable has come into play.

Another example I remember hearing it that in a study of primary school children, it was found those with better grammar had more teeth cavities. Why?

Yakuza

They got sweets as a reward for their performance?

Michael Collins

Yakuza wrote: »

They got sweets as a reward for their performance?

Nope, it's even more basic than that . In fact, if that was the case then it wouldn't be too unfair to say good grammar causes bad teeth because their would be a causal relationship. This example is where there's actually no relationship between the two other than they occur together. Keep in mind the information given in the question is literally all that you should assume, anything else goes!

MathsManiac

Michael Collins wrote: »

Nope, it's even more basic than that . In fact, if that was the case then it wouldn't be too unfair to say good grammar causes bad teeth because their would be a causal relationship. This example is where there's actually no relationship between the two other than they occur together. Keep in mind the information given in the question is literally all that you should assume, anything else goes!

...I would say that it's for the same reason that, among primary school children, reading ability is strongly correlated with shoe size! (Now does reading well make your feet grow, or do big feet help you read better? Hmmm...)