Ah, thesis, how I loathe you and look forward to your completion.
Let me preface this with the fact that I have zero previous education or knowledge of statistical analysis techniques. Yes, I’m sure this is going to be a fun time trying to figure it all out.
Anyway, I have a set of quantitative data from a survey I sent out. I have the standard demographic information (age range, location, gender) and the results from a set of nine five-level likert scale questions. I do have some good information on distribution of answers (i.e. pie graphs and bar charts in Excel), and I know that I should look for the mode for each question rather than mean or median, so there’s that. What I would like to figure out now is if there is a difference in opinions based on their locations and ages. There are only two locations, so hopefully that one isn’t too difficult.
I’m hoping this is something I can do with Excel or an Excel add-on, as the only other data management program I have is NVivo, and I’m not sure it can do quantitative data analysis (hmmm, should double check this, I am just assuming since I’m only using it to code my interview data at this point). I mean, I can buy something else, it’s just a lot of money to spend for something so small on one-time, kwim?
I missed that this is for your thesis. Get yourself to the Math Department (or the Statistics Department, if there is one) and ask a professor for help. Even if they can’t help, they’ll probably be able to direct you to someone who can.
You can do ANOVA since the data is numbers, but it may not be appropriate. If you truly have Likert scale data when the difference between a 1 and 2 response is the same as the difference between a 2 and 3 response, the a difference in means is meaningful. But data is often called “Likert-scale” even when this is not true, but only that 3 is better than 2 which is better than 1.
But I echo that you should talk to a professor who knows this – if not in your department then talk to some one in Stat. If you have no Stat department, you might talk to an Econometrician in the Economics department.
I unfortunately do not have access to a math department or professors. I am doing a working Masters program, which means that I work full time as well as do a Masters full time…primarily via distance learning (I have three, three week residencies every October, otherwise it’s distance coursework). I live about 1400 km’s from my University which is on the coast of B.C. (I’m in Alberta). So, this is a ‘learn on my own’ type of thing. I did remember Khan Academy this evening though, and it looks like they have a bunch of stats tutorials. I think I’ll start going through them in the morning to see what I can learn.
I’ll check out the ANOVA link - that is another that I’ve run in to in my research this evening, so thanks!
To clarify, my survey questions used ‘Strongly Disagree’, ‘Disagree’, ‘Neutral’, ‘Agree’ and ‘Strongly Agree’. I believe this is a properly constructed Likert scale where the difference between each is essentially equal, but correct me if I’m wrong.
I would consider that ordinal data. The chi-square test is appropriate as long as you have a reasonable sample size (say 30 or so), and that’s what I would use.
Also, it seems that there’s a pretty big disagreement as to whether the ‘Strongly Disagree’, ‘Disagree’, ‘Neutral’, ‘Agree’ and ‘Strongly Agree’ choices qualify as interval or ordinal data. If you’re going to use it as interval data, you should cite someone prominent who agrees with that practice.
Not sure what the appropriate analysis is, but if you want a something with more options than Excel with add-ins, you can try OpenStat. It’s free. Google should find it.
A quick search of the manual for “Lickert” found this:
Polytomous DIF Analysis
The purpose of the differential item functioning program is to identify test or attitude items that “perform” differently for two groups - a target group and a reference group. Two procedures are provided and selected on the basis of whether the items are dichotomous (0 and 1 scoring) or consist of multiple categories (e.g. Likert responses ranging from 1 to 5.) The latter case is where the Polytomous DIF Analysis is selected.
I’d agree with the other posters to decide what is appropriate. I don’t know if the above test is what you want. If you have more than two groups, maybe there’s an ANOVA multiple means comparison-like version of this (comparing only two groups would be t-test like). I guess you have two locations to compare and it could do that.
OpenStat has lots of analysis options, including Chi-squared if that’s what you want.
