You should use a paired t-test, the pair consisting of the self esteem and the intelligence variables. You are only performing this on one subset of the data: the tall group.

Remember also to check your assumptions regarding normal distribution of self-esteem and intelligence.

However, I am a bit surprised by your research question - which conclusion will you make from comparing these two variables? It seems a bit like comparing apples to peaches.

Tara Marlena Petzke your answer has contributed to my understanding and I am thankful for that.

Additionally, regarding the conclusions, this is not my actual research question. I just wanted to increase my understanding regarding comparative studies. And your answer has certainly helped.

Raja Ahmed Jamil

The problem is that you should not compare self-esteem and intelligence if they are not standardized. So if you want to use any kind of paired tests you should standardize self-esteem and intelligence before. If these variables have a normal distribution, you may use z-scores (https://en.wikipedia.org/wiki/Standard_score).

But the easier way is just to compare ranks of respondents in self-esteem and ranks in intelligence or you should use some kind of normalization such as quantile normalization for example (https://en.wikipedia.org/wiki/Quantile_normalization).

And you should rank (normalize, standardize) them (tall and short people) as one sample before using paired tests in these two different samples.

Raja Ahmed Jamil

And may be easier way just to compare arithmetic means (or medians) of "tall" and "short" respondents as independent samples by t-test for independent samples or Mann-Whitney U-test. If "taller people" have more self-esteem than “short people” AND statistically significant have lower level of intelligence comparative with “short people” intelligence so looks like "taller people" have higher self-esteem than intelligence comparative to “short people”.