I think this is also a really good section! The summary statistics you're discussing here I would probably include in this section or in your introduction section, rather than in your results section, because they are not so much your analysis so much as your justification for why you think your analysis is worth doing.
Your discussing of the regression analysis is good, and you do a solid job of justifying why you think linear regression is correct and why you are choosing the variables you are (though again, you need to cite the source you're mentioning at least in a footnote).
You also need to justify why AIC and Adjusted R-squared are the metrics by which you are assessing your models (for example, why adjusted R-squared rather than regular R-squared). Also, you should strongly consider breaking your data into a training set and a testing set (and given the amount of data you have maybe also a validation set). You can then assess the models by predicting on the validation set and taking the Mean Squared Error, and then judging the best model you have by predicting on the test set and taking the Mean Squared Error. If you bring all three metrics together, you can make a stronger argument of what is your "best" model.
I think this is also a really good section! The summary statistics you're discussing here I would probably include in this section or in your introduction section, rather than in your results section, because they are not so much your analysis so much as your justification for why you think your analysis is worth doing.
Your discussing of the regression analysis is good, and you do a solid job of justifying why you think linear regression is correct and why you are choosing the variables you are (though again, you need to cite the source you're mentioning at least in a footnote).
You also need to justify why AIC and Adjusted R-squared are the metrics by which you are assessing your models (for example, why adjusted R-squared rather than regular R-squared). Also, you should strongly consider breaking your data into a training set and a testing set (and given the amount of data you have maybe also a validation set). You can then assess the models by predicting on the validation set and taking the Mean Squared Error, and then judging the best model you have by predicting on the test set and taking the Mean Squared Error. If you bring all three metrics together, you can make a stronger argument of what is your "best" model.