You can use your categorical data to split your Likert scales for mean comparisons. For splits of two eg male female you can compare means as an independent t-test. For splits of more than two you can compare means as a Oneway Analysis of Variance. You can of course apply logistic regression. Ensure your Likert scales are normally distributed first and Cronbach's alpha shows reliability level.

if you wish to test categorical variables only can use chi-square tests, if you wish to examine possible relationships of categorical variables further you could use cluster analysis and multiple correspondence analysis (optimal scaling in SPSS)

I suggest you read up on factorial (two-way) ANOVA.  See the link provided below, for example.

http://onlinestatbook.com/2/analysis_of_variance/multiway.html

I would first try crosstabs, than consider recoding the more than two values variable, than try ANOVA.

I would definitely read up on (factorial) ANOVA, as already suggested.

Use Mann Whitney or Kruskall W instead of ANOVA if your data is non-parametric (e.g. non normally distributed, skewed, lots of outliers) because then the means of the dependent variables would give a distorted image.

also make sure to assume that the difference of the values of your Likert scale (used for measure you outcome/dependent varialbe, am I correct?) is equal, otherwise a mean of 2.5 would be meaningless.

Logistic regression is used if you predict categorical outcomes from continuous variables (so the other way around)

I think to compare 2 Likert-type response distributions (1,2,3,4,5, or 1...7), you should use the Mann Whitney U.  It is well understood, has a conventional interpretation of p value, and is the most appropriate test of H0 that median 1 = median 2.

To test H0 among three or more groups (median 1 = median 2 = median n) you should use Kruskal-Wallis 1 way ANOVA.

It is important that you report that  you are in fact testing differences in medians, not means, although t-tests (for the first H0), and basic 1-way ANOVA (for the second H0) should yield similar results if the Likert-type response distributions are approximately normal (or at least not heavily skewed towards 1 or 5 as often happens when measuring strength of agreement or satisfaction, etc.).  How different are the medians and the means?

Chi-square would the referred test between categorical variables. You can run ANOVA for comparing multiple means. Use the Mann Whitney test. I prefer GLM in such situations.

Hello,

If your continuos variable is your outcome I think an ANOVA or a Kruskal - Wallis test would be sufficient. 

As stated before, be very careful if you use Kruskal - Wallis or Mann Whitney because SPSS tends to throw Mean ranks and if I understand correctly you want to compare both groups, in that case your option would be using medians. 

If your outcome variable is categorical (which in this case does not appear so) you could do a logistic regression. 

Your conclusion is sensible but first try to see if you can go linear (use t-student or ANOVA) before going to U-Mann Whitney and KW tests. 

Regards

Would you lie to try Tobit Regression. Stata does it quite well.

Dear Victor;

You can apply two way analysis of variance (gender x profession). In SPSS apply GLM as:

Analyze...General Linear Model...Univariate

See this URL please.

https://www.youtube.com/watch?v=FhcP717lQrk

If there is a significant difference between profession levels you should apply Post Hoc Tests.

Use the average score on the Likert items as the dependent variable  and bring in explicitely your independent variables after coding them  as Dummy Variables (K-1 vectors of ones and zeros) in Multiple Regression analysis. This regression analysis will enable you to test the shift  in intercept  or the change in slope (i.e. if you have to use same equation or different regression equations for male or female etc). Instead you can use Chi Square or ANOVA   but I recommend Multiple Regression Analysis  if you know how to use it because it is more powerful and it is the queen of analysis. 

Good luck

ANOVA (2x3 subjects). you have first to create a new variable for the continuous scale by using the average score on the Likert items and recode your independents variables. 

The Logistic Regression is a reasonable choice, of course.

Another procedure, even more powerful, is the Generalized Linear Model, you can find in the SPSS statistical package, that allows to analyze continuous as well as categorical  scores both for the dipendent and indipendent variables.

For categorical data use crosstabs and chi- square. I would be cautious in computing total score for Likert Scaled questions unless the questionnaire is valid and reliable. Alternatively, you compare mean scores of each questions with gender etc. Use Mann Whitney for 2 groups or Kruskall Wallis for 3 or more groups. There is no reason why you can't do a correlation between categorical and continuous variable using Spearman's. Since you have both categorical and continuous variables a logistic regression is preferred. 

Whenever trying to decide which analysis to use, a key thing to consider is what your research question is. Your data could be analysed in several different ways, depending on your research question. You could use chi square for example, and see if gender (male, female) is dependent on employment status (unemployed, full time etc.). Then again, you might have measured these demographics because you want to control for them in analysing other variables as predictors of an outcome variable. In such a case, using hierarchical multiple regression would make sense, as you could enter your demographic variables in step 1, and your other variables in step 2 to predict your outcome. When you say, “I want to see if the categorical variables are significantly different from each other in terms of means”. Taking this literally, it doesn’t make any sense as the mean of categorical variables is a useless number, e.g. if male coded 0 and female coded 1, with equal numbers of males/females in a sample your mean would be .5 (an illogical number). I’m sure you meant you wanted to compare the mean on a DV of two or more groups (categorical IV). If only comparing two groups, you could use t-test. If more than that, some form of ANOVA. It may help to clarify your research questions/hypotheses and this should help you to be clearer on which analysis you should use.

If you decide to use multiple regression, and want to include your categorical variables, you will as a previous respondent stated, need to convert any categorical variable with more than two groups into a series of dummy variables (see this 10min youtube video for an excellent explanation/demonstration of creating dummy variables and then using them in regression in SPSS, https://www.youtube.com/watch?v=6ceGsClNwzo).