I want to investigate how people who have high job insecurity score on various forms of political participation. So on which form they score higher;
I have job insecurity (low, ,medium,high) in my database, as well as the various forms of political participation.
How do I look at this in SPSS
If you're interested in using only those cases classified as high insecurity (ignoring the other two categories), then a simple omnibus test would be a chi-square goodness of fit test (available in spss as analyze/non-parametric) across the five categories of political participation. That would yield a 2 (yes/no) x 5 table, and the null hypothesis would be that of equal participation rates across the five categories. You could also treat this as a multivariate sign test--see Sheu & Curry (1996), Implementation of nonparametric multivariate statistics with S. Behavior Research Instruments, Methods, & Computers, 28(2), 315-318.
Dear Ann,
I think it depends on what the appropriate measures are for your political participation data. If it is nominal or ordinal then is sounds like a chi-squared test for association is appropriate (although you will need to watch the validity for small expected cell counts). If it is scale then because you will probably be looking for a trend based on your ordinal job insecurity variable I would recommend using the Jonkheere-Tepstra test.
Dear Ann
as Peter mentioned it's important to explain the nature of the dependent variable.
Good luck
Dear Ann,
If your dependent variable is ordinal you can use the bivariate kendall´s tau test to see the direction of the effect.
Hey all, thanks for answering my question.
my dependent variables are 5 forms of political participation, and each form has either a score of 0 or of 1. 0 means no participation and 1 means participation
You canuse Cramers V with each dummy to see, if there is an effect.
Dear Ann,
You can achieve this by using a clustering method, namely a two-step cluster analysis.
Your job insecurity data is ordinal, do you consider the political participation data to be ordinal as well (i.e. left to right). in that case you could do a spearman correlation (if they're not ordinal I would agree with the Kendall's recommendation). Also, are the 5 forms mutually exclusive or can someone have more than one form, which makes the whole thing a bit more complicated.
If you're interested in using only those cases classified as high insecurity (ignoring the other two categories), then a simple omnibus test would be a chi-square goodness of fit test (available in spss as analyze/non-parametric) across the five categories of political participation. That would yield a 2 (yes/no) x 5 table, and the null hypothesis would be that of equal participation rates across the five categories. You could also treat this as a multivariate sign test--see Sheu & Curry (1996), Implementation of nonparametric multivariate statistics with S. Behavior Research Instruments, Methods, & Computers, 28(2), 315-318.
Patrick, 1 form of political participation has for example 3 items in it, another 4 items and a third has 2 items. The remaining 2 have just 1 item, so they are not mutually exclusive.
The 5 forms are: formal, activist, protest, attention, social engagement
(This is purely based on literature and what I am researching btw). But for example activist has 3 items in it, like signing petitions ( done vs not done) and boycotting for example. I then recoded it so a respondent can score 0 (havent participated on any) or 1 ( participated on at least one). I did this based on other research who done it that way, as my teacher suggested to me. In this way I could also perform binary logistic analysis on them all,
Thanks everyone else for the answers to. David I tried your suggestion but have no idea how this works. What do I fill in in the test variable list and expected values and such?
Ann, by testing each separately you run into the issue of multiple testing. In doing so, you are increasing your Odds of finding a relationship. It has been suggested that you adjust your p-value (assuming 95th% CI) that you consider significant by [.05/number of tests). For example, you have 5 items so the minimum p-value that you would consider significant for each test would be [.05/5] p
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS