I have two dataset: population density and case fatality rate. Population density is measured as number of people living in an area divided by that area size in square miles (number of people/area size). Case fatality rate is calculated as total death from an infection divided by total number of infections (total death/total infection). Are population density and case fatality rate intervals?
$\begingroup$
$\endgroup$
6
-
$\begingroup$ Presumably you mean of "interval data type." But could you explain why the answer would be relevant to any kind of data analysis or investigation? $\endgroup$– whuber ♦Jan 10, 2022 at 21:09
-
1$\begingroup$ Both datasets take only non-negative values. $\endgroup$– BruceETJan 10, 2022 at 22:33
-
$\begingroup$ @whuber yes I meant are they interval data type. I want to find if there is a correlation between them. I use population density as the independent value. $\endgroup$– kakaJan 11, 2022 at 21:26
-
$\begingroup$ @BruceET yes. So are they interval data type? $\endgroup$– kakaJan 11, 2022 at 21:26
-
2$\begingroup$ For assessing relationships or correlations, it is superfluous to determine the data type. $\endgroup$– whuber ♦Jan 11, 2022 at 21:42
|
Show 1 more comment
1 Answer
$\begingroup$
$\endgroup$
Both population density and case fatality rate are based on a ratio of a necessarily non-negative value to a necessarily positive value. That's the definition of a ratio-scale measurement.
As comments have pointed out, you don't need to distinguish interval from ratio data types to do correlations.* Also, correlations are generally considered to be symmetric between 2 variables, without a defined "independent" variable. That distinction becomes important when you are doing regression.
*I don't find the "interval" measurement type as defined by Stevens to be very helpful, except to provide a warning about taking ratios. I suspect that most users of this site would lean toward the Mosteller/Tukey typology, perhaps extending it in particular cases with things like cyclic data.
