I am exploring the readability of court decisions in 5 countries. I have a quantitative readability measure (continuous variable, my dependent variable) for each decision - about 50 decisions per country for a total of about 250 lines of data. I want to assess whether independent variables applicable to each country (like "percentage of the population in that country that doesn't finish secondary school" and "mean number of post-secondary degrees per judge in the country" and "mean number of law professors per judge in the country", etc) are correlated with readability levels of court decisions in those countries. These independent variables are all of the same value for each of the 50-ish decisions within a given country. Can I use Pearson? If not, is there some other way to show how the independent variables are/are not related to the dependent variable? ANOVA? (There are only 5 discreet values for each of the independent variables...one value for all decisions within a given country, times 5 countries.
Yes, you may use PPMC but make sure you have to transform the continuous data by computing the means of these variables (Compute variables) for each country before you run PPMC. The variable transformation is to change the data to parametric data. You can even go further and rank them based on the influence level of IV on DV in each country and see which country is flagged up (focal point) and which is not
Dear Mike Madden ,
Please first advise if your data strictly cross-sectional or cross-sectional timeseries (panel dataset - gathered across countries over time - eg a particular one period/point in time or over the years). That information, together with the your research objective and measurement levels of your DV and IVs you have already shared, will help in making informed suggestions of appropriate analysis methods.
Hello MIke,
If you were surveying results from many countries, then correlating the country-level values for variables such as # law professors per judge with the averaged value of readability would be feasible. With only five countries, you'll not have a great deal of precision in any correlation you compute.
If it were possible to get values for these variables at a lower level than country-wide, you'd probably be able to proceed with the 5 country set-up.
Good luck with your work.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA