I am collecting data from a group of about 189 police officers on a 5-point Likert-Scale (measuring how much the police officers accept rape myths). I have no idea which tests to use to analyse my data. In the end I want to divide the group into male officers versus female officers and compare their acceptance of rape myths to determine if there is a difference between the two groups. I will also gather other geographical data (e.g. years of service), but I have no idea how that can be valuable.
$\begingroup$
$\endgroup$
6
-
$\begingroup$ By "geographical", do you mean demographical? What questions do you want to answer, only if men & women differ? $\endgroup$– gung - Reinstate MonicaCommented Nov 27, 2014 at 13:17
-
$\begingroup$ Research question:Is there a difference between female and male police officers’ male rape myth acceptance in South Africa?The MRMS is a 22-item questionnaire which measures participants’ acceptance of untrue or stereotypical beliefs regarding male rape (I'll be using the MRMS).Participants respond to the questions by choosing an option from a 6-point Likert scale ranging from strongly disagree (1) to strongly agree (6).The following demographic data will be collected: Gender, age, ethnicity, years of service, as well as place of residence. I have never analysed data, thus I am very lost. $\endgroup$– newBCommented Nov 27, 2014 at 13:28
-
$\begingroup$ If you truly have no idea, then you must not have read any of the Related links on the right side of this page. That's a place to start. If you are struggling with a choice between two or more alternatives, please say something about that. $\endgroup$– rolando2Commented Nov 27, 2014 at 13:43
-
1$\begingroup$ I have read so many websites and I have gathered that ordinal data (Likert-scale data as far as I gather) can't yield mean values - something to do with the distance between the Likert-scale options.I have read that calculating the median and Inter-Quartile Range (IQR) for the MRMS questionnaire will work, but I don't really know what that means.Also, it seems that a Mann-Whitney U-test can be used to examine whether rape myth acceptance differs between male versus female police officers, but I am very unsure... $\endgroup$– newBCommented Nov 27, 2014 at 13:46
-
$\begingroup$ If you really don't know anything about the issues involved here, you should hire a statistical consultant. CV is good for getting answers to focused questions when you are only missing one piece of the puzzle. When you are missing all the pieces of the puzzle, you will have a very hard time & CV won't be enough help. $\endgroup$– gung - Reinstate MonicaCommented Nov 27, 2014 at 15:13
|
Show 1 more comment
2 Answers
$\begingroup$
$\endgroup$
4
It depends on whether you have a single Likert scale question(which, given your research question, seems unlikely) or multiple scales. If the latter, you then have to decide how you want to combine them.
Although, technically, adding Likert scale questions is invalid (because adding numbers assumes they are at least interval scaled) people do it all the time. It is also possible to do factor analysis of Likert scale items, although some care must be taken. See e.g Joreskog and Sorbom
answered Nov 27, 2014 at 13:50
-
$\begingroup$ My questionnaire has 22 questions all measuring the same thing - male rape myth acceptance.I am so new to this that I had to look up "combining multiple likert-scales" and "interval scaled" so I take a while to respond. $\endgroup$– newBCommented Nov 27, 2014 at 13:55
-
$\begingroup$ If you have $19 and 2 hours to spend I highly recommend Paul Spector's little book at sagepub.com/books/Book3653 . $\endgroup$– rolando2Commented Nov 27, 2014 at 13:59
-
$\begingroup$ Thank you so much Ronaldo2. ps. I did read the Related articles and I gathered some information, but (to be honest) most of it was very confusing. $\endgroup$– newBCommented Nov 27, 2014 at 14:02
-
$\begingroup$ If you are truly a NewB and had to look up interval scaled, that article is going to be heavy going. You may want to just add the items, or else seek help from an expert. $\endgroup$ Commented Nov 27, 2014 at 16:00
$\begingroup$
$\endgroup$
4
Other than the likert question, If you have all your variables as categorical e.g. Region(East, West, North, South), Education(Some College, College Degree, Masters Degree, Phd) and so on, you can use MANOVA technique.
"Rape as a myth" as dependent variable. Gender(Male/Female) as fixed factor
Looking at your question, you have "Age" as a continuous variable. you can actually segment age into diffeent defined groups to make it categorical variable and make it useful for MOANOVA. For example you can have age categories as "18 to 25", "26 to 30", "30 to 35" and so on. So now if a respondent stated their age as 23, it will fall under "18 to 25".
This is an youtube SPSS tutorial on how to do MANOVA and how to interpret the output
answered Feb 15, 2015 at 7:01
-
$\begingroup$ Grouping ages unnecessarily loses information, though it may be the case the original questionnaire used such intervals. Even so, treating age as categorical rather than, at least, ordinal would ignore the ordered nature of ages. $\endgroup$ Commented Feb 15, 2015 at 8:27
-
$\begingroup$ @Silverfish Care to elaborate with an example? I have seen and used grouping frequently in market research. If we are not sure how t start with it, we can do visual binning in SPSS which gives us idea about possible ranges. $\endgroup$ Commented Feb 15, 2015 at 11:09
-
$\begingroup$ See the thread on statistical sins for a discussion of the extreme case of this (dichotomization). There's good discussion in the comments in this thread though sadly it doesn't have a good answer at the moment. $\endgroup$ Commented Feb 18, 2015 at 0:11
-
$\begingroup$ Even more comprehensive, see the notes by Frank Harrell $\endgroup$ Commented Feb 18, 2015 at 17:38