0
$\begingroup$

I have 257 variables, which influence behaviour altruism of humans. They can be assigned to the 3 following criteria: internal, external or selection. I check the influence of the individual variable on altruism, further I need to check the influence of all variables, which are part of internal. Can I just standardise these variables and then take the mean of them and consider this as an index?

$\endgroup$
2
  • $\begingroup$ Can you say more about your situation, your data and your goals? I suspect a lot is being assumed here. $\endgroup$ – gung - Reinstate Monica Jul 17 '16 at 2:24
  • 1
    $\begingroup$ Sure: you can apply literally any mathematical formula you like and call it an "index." The crux of the matter, though, is whether it bears any relationship at all altruism. To do that, one would think you ought to analyze the possible relationships between the variables and altruism itself. Since standardization doesn't tell us anything at all about relationships--it's strictly a univariate process--it's difficult to see how it would accomplish anything useful. $\endgroup$ – whuber Jul 17 '16 at 14:09
1
$\begingroup$

As Gung said, we should know a lot more about your situation to give a really useful specific answer. Anyway, I'll try to give a general one.

The altruistic behaviour you are interested in may be influenced by some of your 257 variables. You need to know which variables do influence altruism, and therefore you should resort to tools like regression or ANOVA that could tell you which variables have a significant influence on altruism. Please notice that such tools check influence of every variable - at least, every variable you tryed to include in your analysis.

However, if you use an index (which in your proposal is just a single linear combination of your variables out of the infinite amount of linear combinations you could use), you will lose a lot of information, and the influence of single variables can be cancelled by the influence of other variables or just get lost in the sum.

Just to put an invented example with two variables: Let's imagine that people with more body weight are more altruistic, and that tall people is less altruistic that short people. If you ran a regression analysis you would find significant positive correlation between body weight and altruism and negative correlation between height and altruism - and that would be what you were searching for. Although, if instead of using body weight and height you use an index made of the (standardized) sum of both, the effects might be cancelled - that is, the most altruistic people (short people with overweight) will get about the same index value as the less altruistic ones (tall slim people), and therefore you wouldn't be likely to find any correlation between your index and altruism.

In summary, if you have a lot of variables, use tools to tell which of them have influence. You may design a useful index if you have good reasons to build it in a particular way, but if you just take an arbitrary index like the standardized sum of all variables, you are just losing information - a lot of information.

$\endgroup$
2
  • 1
    $\begingroup$ Sorry for the late reply and thank you very much for your good example. I collected data (400 observations) in 40 countries where I gave the person something and checked the reaction. This reaction could be positive (1) or negative (0). I am now interested how this behaviour changes across country based on socioeconomic factors. I analysed the dahlström Dataset link to check on theoretical connection. Would it I would now like to create a simple model where I assign the variables to either one of the three criterions. $\endgroup$ – user123541 Aug 5 '16 at 16:53
  • $\begingroup$ Now education has a positive influence but gini coefficient has a negative influence. So taking your example with hight and weight, would it be best to analyse the way of correlation first and then adapt them so the effect is always to the same direction, so that e.g. high values are always more altruistic, so a small altruistic person gets a score that is negatively correlated to the actual hight, so that they don't cancel out? Or does this cause other problems @pere $\endgroup$ – user123541 Aug 5 '16 at 16:59

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.