First, it seems to me that school degres and nationality are quantitative, i.e.it is possible to identify them clearly, to makse statistiques, etc.  I beleive that those three varaibles can be quantitative.

At this juncture, in agreement with the answer above, I can only point out that the variables 'school degree level' and 'nationality' are quantitative; the first, 'school degree level' in terms of it being a numeric or alpha-numeric combination ( e. g.1/2/3 or A1/B1 etc.); the second, 'nationality' in terms of it probably being expressed for analysis with your other data, as a number.  (You seem to be intending a quantitative study).

It seems that you first ask for scales and second appropriate analysis techniques. If school degree is linked to persons it can be seen as a ordinal scale which is not anymore really quantitative. But as you are able to define ranks for salaries as well as the degree level you can calculate a Spearman rank correlation. Nationality is to me a nominal and by this a qualitative variable. If you have N different nationalities in your data you can transfer them to N-1 binary variables (if all N-1 binary are 0 the nationality is the last "N" one - be careful not to use N binary as otherwise you create a multicolinearity in your dataset of independent variables). Of course you can use regression with binary data - by this you will see which nationalities have a significant positive influence to salaries. Finally - which is statistically not really proper you also can treat an ordinal variable as quantitative if there is not a to big difference between Pearson and Spearman correlation. But doing this you should be carful with interpreting test statistics. And very finally you also can code a variable degrees with D degree levels as D-1 binary variables. 

Dear Risqia

Yes, their is rank correlation and logistic regression. Pl check, if it serves your purpose.

Regards

It sounds like your goal is to predict salary from a set of variables. This is often done with multiply regression. Salary (and the residuals after model fitting) is usually positively skewed so a transformation is often used (e.g., logging them), but this will depend on your data and the purpose of the analyses. Qualitative variables like nationality (I assume that you are not assuming an underlying metric for this, so it is qualitative) are usually included in regressions by calculating contrasts for them. The simplest method would be treating your largest category as the reference category and then have dummy variables for others, but the choice of contrasts will depend on the purpose of your research and your data.

Thank you for all of your answers, I really appreciated it all.. it helps me a lot.. separating between categorical and continuous variable makes it easier.. also I assumed categorical as numeric with a control variable so I can calculate the correlation..