Close Home Forum Sign up / Log in

SPSS- any ideas?

B

Please excuse how utterly thick I am, but I have a hate-hate relationship with SPSS and really don't understand it all.

Okay, so I've got to combine several variables into a new variable. I've done that bit somehow, but now I've got to see whether the new variable (a scale variable) is related to social class.

What is bugging me is that I can't work out whether the social class variable is nominal or ordinal :$ there are five responses it can be, ranging from middle to working to poor. Any ideas?

Based on this, I have tried to do both a box plot (if it were nominal) as well as Spearman's rho/Pearson's/Kendall (if it is ordinal).

But firstly, when I try to do a box plot no results come up, like nothing appears on the graph, there are no box plots.

And when I do any of Spearmans/Pearson/Kendall either nothing comes up for the correlation coefficients or it says "cannot be computed because at least one of the variables are computed". Eh?

Any ideas are very much appreciated, I'm having an absolute nightmare!

button

J

I would say that social class is ordinal data because you can rank it (middle class being highest status, poor being lowest).

Can't help you on the other factors as I no longer use SPSS (I've turned qualitative) and would need to look at the data to work it out!

Good luck!

J

I agree that is is ordinal. I'm not a huge expert on SPSS but a colleague has recommended a book by Julie Pallant with she said taught her everything she knows (and she is fab at it) - maybe get that from your library

B

Thanks for your help, I'm glad that you both think it's ordinal as that was my original thought, but because the stats wouldn't work I started questioning myself! Anyway, I'm still stuck on it, and I've decided I need to use Spearman's rho, but it doesn't give me the correlation coefficient because N=0, when it should be lots of participants...any ideas anyone?

i never mess around with ordinal/nominal. I just chuck everything in as scale data! there is page in andy field somewhere that says there are arguments for using likert type data e.g. strongly agree - strongly disagree as scale data.

I would leave it as scale data but make sure you interpret it correctly e.g. moving from 6 to 5 means moving from poor to working class.

Or maybe you need to dummy code it?

K

If you are analysing a scale variable (i.e. age) with an ordinal variable like social class, then you can use linear regression to see if there is a correlation and how strong it is.  You wouldn't need to recode the variable into a dummy (although you could) as long as the ordinal variable isn't coded like a dogs breakfast.

So, you would want 1 to 5 say, 1 being lowest class and 5 being highest class (the other way around would work too but in a sense counter-intuitive because if you had a negative correlation it would be towards the higher class).

If you go into analyse/regression/linear and put your scaled variable as your dependent variable and class as your independent you should get an R value, and R-squared values and B and beta scores which will tell you how strongly social class affects your scale variable.

Send me a PM if this completely confuses you.

14244