28th March 2014 at 12:26 pm #1106Kassa Teshager AlemuMember
Dear All, I am currently Analyzing a data collected by likert scale items. I want to quantify in the form of index or mean value. is there a simple way to do it in SPSS? Again how reliable the data could be to infer..28th March 2014 at 4:00 pm #1111Dave CollingridgeParticipant
One way to aggregate Likert-type scales is to sum individuals’ values across questions and then divide by the number of questions you combined. You should run a principal components analysis first to find out whether questions load onto the same components. Questions loading onto the same components can be combined. Questions that do not load onto components can be analyzed separately with non-parametric statistics.29th March 2014 at 12:06 pm #1110Stephen GorardParticipant
My advice would be – don’t. Ideally you would have worked out what you wanted to do before collecting data so that you could ensure that what you collected was the right stuff in the right format for your research question(s).
What you seem to be proposing has several problems. First, the responses you have are ordinal classifications not scores. Second, averaging across items (even if they were scores) is an error. If the items correlate highly (+/-1.0 or near) then it shows you have asked the same question repeatedly (wasting respondent time). If they correlate little (+/-0.7 or lower, meaning that over half of the variance is unique to each item), then adding them would be like adding a height to a maths test score. If the error in each item is random, the only reason really for asking the question several times, then repeating the question must reduce the accuracy of the average responses.9th April 2014 at 12:08 am #1109Alexander Essien TimothyParticipant
Hi Massa, I suppose your scale has up to five categories. You probably have the following responses: strongly agree, agree, disagree, strongly disagree, undecided. For positive statements such as: ‘I wash my hands before eating,’ you’ll score the response thus – SA -5; A – 4; D – 3; SD- 2; UD -1. Then sum all the scores on the responses pertaining to the particular variable for each respondent. The sum of scores of the respondents on the variable can then be calculated and divided by the total number of respondents to have your mean.
I hope this helps.
Alex9th April 2014 at 5:14 am #1108
You are using the most standradised tool, or a scale, likert scale. Your data should be highly acurate. then why do you again want to requantify. For each sample take the total score aggeregate and divided by n, that is how you get mean score.
As per the rule of thumb, in this scale , the 5 point, ranging from most to least agree. SA,A,UN,D,SD. Positive 5,4 3,2,1 and negative statements reverse1,2,3,4,5. So now when youb have total the scores of all the item you get exact score.(remebr positive+negative gets nullified). Then you take the mean.
This is the standard method in quantitaive research, which scholars here do.
HOD,Research department,PSG,Coimbatore,TamilNadu,India9th April 2014 at 12:01 pm #1107Kassa Teshager AlemuMember
thank you very much for your inputs. I have managed to do it accordingly.
- You must be logged in to reply to this topic.