Statistics – Cronbach’s alpha
Hi, after my webinar in beginning of July about factor analysis and creating index variables of variables that is correlated to each other, I think it’s worth telling you about a validation method before creating index variables. Index variable could be a sum/mean of some questions (variables) that is correlated to each other, ex satisfaction index.
The method to check correlations between a couple of variables at the same time, is called Cronbach’s Alpha. The alpha value you will get is between 0 and 1 and is like a “average correlations measure” for a couple of variables – often questions from a questionnaire. It should be above 0.7 – a common treasure value.
A rule is: Just use variables with positive correlations. Also be aware of that the alpha-value will be higher if you have more variables, so if you have more than 5 variables and one of them has a bad correlation to the other – the alpha value could be higher anyway.
From the factor analysis I found out that these 3 variables within the red oval, belongs to each other (correlated to each other):
So as they all is positive correlated I want to measure the Cronbach’s alpha, to see how strong all 3 correlates to each other at the same time, so I put these 3 variables (questions) into the right box at “items” in the command (reliability analysis):
I’m not so happy, as the alpha value (0,62) is not above 0,70. What happen if I add the 4th variable that is positively correlated to these 3, if you look at the factor analysis matrix above (“it’s important to keep the flow in the traffic…”).
So it’s closer to 0,7, but I cannot be sure that this is because of a good correlation among all 4, as I also know that the alpha value gets higher when you add an extra variable.
So the reason why the alpha value is not so good as I hoped for, is because of the correlations among these variables is not very high. I will talk more about correlations next week.
Here is an example of 3 other variables that has much higher correlations, and what happens to the Cronbachs alpha:
I got a good result, and could now create an index between these 3 variables.
Help: If you double click the result table and then right click on Cronbach’s Alpha you will get a yellow box with information:
Let’s show you the correlations of these variables next week!