Reliability Analysis

okay this is the lesson in spss pasw on how to perform an alpha reliability analysis this is an analysis you run any time you have scaled data you want to make sure that the items are they correlate with each other they hang together as they should and it's called a reliability analysis so anytime you have a scale and you want to form an average or a composite of different items from that scale you should really be running an alpha reliability analysis so in this case we have a couple columns our participant ID and participants X and then we have five columns representing five different food items questions on if one looks like a five-point scale assessing people's mood and let's see my goal is to form a composite of these items i I don't want to have to analyze these separately my goal is to simply form an average and make sure that and just run the analysis on the mood generally maybe I want to see if men and women differ in their overall mood but before I form that average I really should do what's called the reliability analysis to see are these items all actually measuring are they all tapping the same construct that I'm calling mood so we're going to do is we're going to have to analyze we're going to go down to scale and we're going to go to reliability analysis this is the option we use when we have a scale and we want to see if the items are reliable with each other if they hang together so when I click on this and we'll reset and what we want to do is we want to bring over all of the items that are relevant to our scale in this case mood one two three four and five we're going to leave it set on alpha there are some other options for advanced users but in general alpha is the most common we're going to click on statistics and we're going to click on the following three things under descriptives we want item descriptives scale descriptives and scale if item deleted we're always going to click these three things and we'll ignore everything else we'll then hit continue and we'll then hit okay so we get a bunch of windows in our output how many cases were present there were 18 with 18 participants either no missing data we then come down here and what we're looking for is this number here the cronbach's alpha and ideally we want this number to be above 0.7 positive 0.7 it has five items we want this do two we want this to be above 0.7 and in this case it's negative in fact so that's a very bad it means that there's something wrong with our scale um if we go down we have our item statistics that lets us know what is the mean for item 1 for item 2 for item 3 if we look at this it looks like the mean for item 4 is definitely higher than the other means maybe that might be the problem item total statistics lets us know what happens if you delete a particular item from the scale so if you delete scale item 1 the new convex alpha is negative point 6 9 so it's even worse and if you delete 2 or 3 or 5 it gets considerably worse but if you delete item 4 from the scale your convex alpha now is positive point 8 5 meaning that the scale would be much much better if we only use items 1 2 3 & 4 if we in fact go up to analyze scale reliability analysis and take out item 4 and rerun this analysis our new convex alpha is in fact 0.85 now one reason that the scale may be bad and make atom 4 may have been bad but it's also possible that I didn't reverse code that question let's say items 1 2 3 & 5 higher numbers mean more happiness but for item 4 higher numbers mean less happiness less positive mood in this case we would want to reverse code this item a way of reverse coding is we can go to transform compute variable we're going to take item mood item 4 and all we're going to do is we're going to do 6 minus mood item 4 and we'll call this recoded now the reason I do six - mood for is that every single number is is one through five which means that if I want a five to actually be a one I could go into recode into different variable and do it that way or if I do six - mood then a five becomes a 1 a 4 becomes a 2 3 becomes a 3 so it were essentially reversed coding automatically if I hit ok and I go back here and now have my recoded mood instead of a 5 it's a 1 instead of a 1 it's a 5 instead of a 4 - 2 and if I now run the reliability analysis and I put back in recoded mood for and I hit OK my cronbach's alpha is now 0.88 7 this indicates to me that this is a good scale then I can in fact form a composite of this scale that is how you perform a cronbach's alpha analysis in SPSS slash pasw

Channel: bernstmj

Views: 161,888

Rating: 4.9041915 out of 5

Keywords: Statistics, Psychology, Academia, Math

Id: qcJNXg6HosE

Length: 5min 17sec (317 seconds)

Published: Tue Jun 28 2011