I'm running an analysis about a dataset with three variables: Income, number of residents per residence and percentage of income spent on rent (response variable). What descriptive statistics should I use?

  • Percentage of income spent on rent (continous (0,1))
  • Income continous
  • Number of residents per residence (discrete (1,2,3,4,5,6,7))

What I think:

  • Scatter plot (variable response x independent variables)
  • Histogram of income (with normal curve)
  • Barplot of number of residents
  • Mean, sd, etc for income

Make sense I do a boxplot of income percent spent with rent by number of residents? I ask it because I have only two observations where the number of residents is 7.

What more I can do?

EDIT: Is there any problem in using a histogram for a variable that represents some proportion? I want to show that this variable is asymmetric and heteroscedastic

  • $\begingroup$ What you do ought to be motivated by the purpose of the analysis. Why are you doing it? $\endgroup$ – whuber Apr 5 '17 at 23:12
  • $\begingroup$ @whuber I want to check if there is an association between the percentage of income spent on rent and the income and number of residents. Later I will do a beta regression model for it. $\endgroup$ – user72621 Apr 5 '17 at 23:14

I personally like violin plots. They help you see the distribution for each category.

The plots you suggest are all good. You can maybe add the violin plots vs number of residents and maybe group together 5+ residents (if you don't have that many observations).

Furthermore, you can check for correlations between the independent variables again using violin plots or boxplots and maybe do a spearman's correlation test.

  • $\begingroup$ +1 I just learned about violin plots. I don't know how you would use them to do Spearman's correlation test. $\endgroup$ – Michael R. Chernick Apr 5 '17 at 23:46
  • $\begingroup$ This is separate from the violin plots. Just to enhance your exploratory analysis with some numbers/statistics to see if your independent variables are correlated $\endgroup$ – Vasilis Vasileiou Apr 5 '17 at 23:49
  • $\begingroup$ @VaslisVasileiou I see that as inferential and not exploratory. . $\endgroup$ – Michael R. Chernick Apr 5 '17 at 23:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy