Hi,

Im a masters student. Currently working on my final thesis. There i have a likert scale questions about e-commerce related costs(15 cost variables). If cost are highly significant respondents may strongly agree with the question. Likewise there are 1-5 rating where 1=strongly disagree, 2= disagree, 3= indifference, 4=agree and 5=strongly agree. I need to set an hypothesis as the costs are highly significant in my sample.

Can i use analysis where it calculate the mean value of the response with Standard deviation? would that analysis provide a good interpretation to prove my Hypothesis? If not what kind of Statical method should I use?

Please help me on this issue asap!
thanks very much
Eranga

You have ordinal level data; it would not be appropriate to calculate means and standard deviations. Calculate the interquartile ranges and medians. I'm not sure what you mean by costs being highly significant. How are your costs highly significant. Are you trying to compare groups? So cost A vs cost B? Very broadly, because I'm not quite sure what you're trying to do, the test I would think you need is the Sign Rank test (there is one for paired and unpaired data).

thats funny Wal - in my field its perfectly acceptable to tread 1-5 likert scales as scale data in spss and use means and SDs. As long as you could consider the jump between each category as equal.

======= Date Modified 09 Dec 2010 16:44:45 =======
Hi

Thanks for quick reply. When the mean values is higher the significant level of cost is high. For instant i asked about Hardware cost, mean value of that data column is 4.08 ( Sum of response values/ N). So that cost significant level is high. I dont want to compare groups. I have a hypothesis as " E-commerce related Costs are highly significant (highly important)"

is that make a sense
thanks
Eranga

It's probably a profession-specific thing, Sneaks. I admit, it's a lot easier if you handle the data as parametric (and it's just me in thesis-defence mode).

erm, do you mean 'statistically significant'?? you have to take into account what you're comparing it against - is 4.08 might be a high score in one population, but might be a really low score in another.

So, can't I make a Hypothesis as "Significant levels of costs are high among my sample"? using this data?

thanks
Eranga

I'm not sure if I'm understanding right, but think you want to know whether costs are highly important in your sample (ie people respond with 4/5).

I'm kinda musing so don't know how helpful any of this is.......

You could do a chi sqaure test of the number selecting each category against the nul hypothesis that scores are equally spread accross all categories.....except this null hypothesis assumes that people are equally likely to selct all values on the scale, when people are actually less likely to select the extreme values. I suppose you could get ound this by collapsing 4 and 5 together and 1 and 2 together.

Another idea could be to compare scores accross the 15 items on questionnaire to see which are most important. But I don't know how that fits your hypotheses.

I'm not sure I'd use means and medians with the data you have as it is ordinal rather than scale, and you can't really argue that the difference between strongly disagree and disagree is the same as the difference between disagree and indifferent for example.

:-) thats fine. Thanks all for the quick help on this tread!