0
$\begingroup$

So I have 4 columns of data, shown below. The first column is the independent variable and the next 3 columns are the values of dependent variables.

1    223064 222724  222672
2    220066 219400  219298
3    217105 216128  215978
4    214105 212906  212709
5    211295 209732  209492

I would like to calculate the relative difference among the 3 dependent variables with respect to the variation in the independent variable, and the output should be a single value for each row. To clarify, the 3 (dependent variable) columns are values for the same event but for different times (independent variable) and they are diverging as we go downwards. Basically, I am trying to show the divergence of the data.

One way I can think of it is to calculate the difference among the columns themselves, ABS(1-2), ABS(2-3), ABS(3-1). Then Calculate the Mean and Standard Deviation of the 3 differences.

Is this a good way to do it? or, is there any other method to show the relative differences among the columns beside the above one suggested?

$\endgroup$
1
$\begingroup$

In general, standardized measure of difference can be called "effect size". Here is a nice post discussing some different effect sizes for 3 different groups (columns) of data. I would consider the general class of answers to your question "effect sizes for one-way ANOVA".

A common one (described in the link) is $\eta^2$. Mentally, you can consider this as finding out what fraction of variation of all of your data is due to group differences.

$\endgroup$
  • $\begingroup$ I did one-way ANOVA by following this tutorial, and it gave me F<Fcrit. $\endgroup$ – ShadowWarrior Jun 14 '17 at 7:33
  • $\begingroup$ Sure. The F value of a one way ANOVA could itself be useful. If you are familiar with the ANOVA process, $\eta^2$ is just the SSR (or SSB) divided by the SST. I should add that $F<F_{crit}$ gets into hypothesis testing, which moves a bit away from your original question. Measures of effect size, in general, will be more true to what you have asked. $\endgroup$ – mfloren Jun 14 '17 at 13:38
  • $\begingroup$ I might add that these data are corresponding to 1 independent variable (1 additional column left of the 3 columns of data above). How do I show the effect of the change in the independent variable on the dependent variables? Apologies, as I am new to statistics. I have updated the OP above. $\endgroup$ – ShadowWarrior Jun 17 '17 at 16:37
  • $\begingroup$ In this case (with three outcomes for different categories of an IV) you may change to a "MANOVA" approach (specifying the multiple dependent variables). Again, just check out some stuff for "MANOVA effect sizes"! $\endgroup$ – mfloren Jun 17 '17 at 16:41

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.