hepl re rank order anlysis

pollydollydoda

apologies another cry for help
I am currently attempting to write a thesis, I have my data of 56 questionnaires,
from one group, three parts- demographics- 20 question knowledge test and 2 rank order q's where the participants rated from 1 to 10 the perceived importance of certain activities in their role !
I have downloaded SPSS and have done my descriptive analysis of the first two parts(I think) but I've no idea what to do with the rank order q's do I document the mean of all 56 answers for each activity or do i add all the answers and give the sum total ( really hadn't a clue what implications this would have when i picked my questioning)
also I want to look at the score in the knowledge test and its relationship with age so i did a pearsons correlation test am I right ???? i don't live near my college and don't know anyone near me with stat knowledge !!!!!
thanks

Ostrom

pollydollydoda wrote: »

apologies another cry for help
I am currently attempting to write a thesis, I have my data of 56 questionnaires,
from one group, three parts- demographics- 20 question knowledge test and 2 rank order q's where the participants rated from 1 to 10 the perceived importance of certain activities in their role !
I have downloaded SPSS and have done my descriptive analysis of the first two parts(I think) but I've no idea what to do with the rank order q's do I document the mean of all 56 answers for each activity or do i add all the answers and give the sum total ( really hadn't a clue what implications this would have when i picked my questioning)
also I want to look at the score in the knowledge test and its relationship with age so i did a pearsons correlation test am I right ???? i don't live near my college and don't know anyone near me with stat knowledge !!!!!
thanks

For the rank-order questions it depends on the nature of the variable. Quite often such questions can be treated as interval, even through strictly they are ordinal, in which case you may report the mean, and use pearson's coefficient.

There is no need to add all of the responses together, although I'm not sure what you mean by this. Are you trying to make a new variable from the combined rank order questions? Based on your post, it doesn't sound as if you are, so there is no need to add (although this is a fairly straightforward procedure).

You should report the median in your descriptives also just to be safe, as this is typically required for ordinal measures. You can find this at Analyze---Descriptive Statistics---Explore. Transfer the appropriate variables into the output box and it will give you a range of descriptive figures, including mean and median.

As for the correlation, can you be more specific as to how the rank order questions and age were measured? Did you record age as a raw number, or ask them to tick a box (i.e 16-20, 21-25)? Assuming age was recorded as a raw score, and the rank order questions are equal interval, you can use pearson's without coming to any real harm. Pearson's is commonly used on other ordinal measures such as likert scales (even though strictly it shouldn't), but if you are in doubt, use spearman's coefficient which is appropriate for ordinal data. Again, this depends on how the rank-order variables were measured, but both coefficients are available from the same tab (Analyze---Correlate---Bivariate).

The rule of thumb in these cases is to work from the lower level of measurement (i.e. if one variable is interval and the other ordinal, use the ordinal test).

pollydollydoda

thank you for replying.
Apologies if I sound clueless (I am regarding this)
for example the rank order questions relate to activities of the job and participants are asked to rate their importance to their role e.g. paperwork, advocacy, problem solving . so I have ten activites with a possible 1-10 score .
with regards to the correlation Q- each participant scored between 4 and 15 out of 20 and I just want to see if the older participants scored higher.Age was documented in groups 15-25, 26-35 etc..

Ostrom

In this case spearman's is appropriate, although if the ages are clumped together in one or two groups it might not be suitable. I really would need to see your dataset to advise, since you might not have much variation across age given that the categories are quite wide (i.e. ten year intervals). If you have a few cases in each age category it should be ok.