I need some advice about my statistical analysis tool i have used for my research. It is for a university project that I am doing and so far i have had no luck in finding the answer to my question.
1) my research is to identify what the physicians current expectations are and how that compares to their actual experience of the role of clinical pharmacists in general practice. 2) i used a 2-part questionnaire method using a 5-point likert scale to identify the physicians current expectations (questionnaire 1) and actual experience (questionnaire -2) of the role of clinical pharmacist in general practice. There are 22-statements and the questions are the same for both parts. 3) all the physicians are expected to fill both parts of the questionnaire therefore it is a single group/population. 3) i was wondering if the wilcoxen-signed rank test/ wilcoxen-t test is a reliable method for evaluating if there is a statistically significant difference between current expectations and actual experience. I understand that likert scales use ordinal data and non-parametric analysis is more accurate. although ordinal data is categorical in some circumstances it can be treated as continuous data. In analysing the data i proposed.... - likert = strongly disagree(1), disagree(2), undecided(3), agree(4), strongly agree (5) - population number is 377 - 2-part questionnaire - random sampling technique. - results tabulate the number of participants that strongly disagree, diagree, undecided, agree, strongly agree. - the median is calculated for each of the above categories. (this is done for both current expectations and current experiences) - the wilcoxen-signed rank test compares the experience and expectation for each category separately 4) my question - is this the correct analytical tool? if not then what test/analytical tool should i be using and why? 5) finally if it is not possible to compare the data, how could i rephrase my research question so that i do not have to start all over again. I am currently at a standstill with this and i spoke to my lecturer but she was uncertain of my method
You can speak to the statistics lecturer. But a Chi squared test would be appropriate to test a Likehurt scale. But Wilcoxon will work too. There are no rules!
Because you have so many measures, you would need to run a large number of separate tests for your pairs of variables. A common alternative would be to find out if your items form a scale or set of scales, which would greatly simply your analyses. Because there have been so many questions here about how to create scales from Likert-scored items, I have compiled a set of resources on this topic:
https://www.researchgate.net/post/How_can_one_help_people_create_simple_scales_from_Likert-scored_items
thanku for replying
i was Thinking ....
1) because I have 22 statement For both expectations vs experience - on 1-5 point Likert scale. I could work out the sum of the frequencies and then median for each statement n compare the results. I understand I would have to do this multiple times for each statement
or
2) I could work out the sum of the frequencies and then the median for strongly disagree, disagree, undecided, agree, strongly agree for each measure and Compare with the same measures for the second questionnaire which would probably be easier
i making the assumption that the Likert scale is continuous variable based on justification
what I am trying to do is evaluate if there is a statistical difference between expecitations experiences.
i am not so sure if rather than using the median I could use averages?
problem is the results are likely to be skewed therefore I thought working out the median is better?
- Badges
- Science topic
- Similar topics
- Correction