search

DS4B-ANU / help

this is where to ask for help with your coding in BIOL3207/6207
MIT License
2 stars 0 forks source link

new questions #10

Open Lizhaozhe123 opened 4 months ago

Lizhaozhe123 commented 4 months ago

Sorry for bothering you ! there are several issues I really want to figure out about week 8 workshops.

1.When we predict the relationship between genome size and wood density, why do we come to the conclusion：our result could still be down to genome size and/or wood density having strong correlations with one or more covariates?(where you can click the accounting for covariates to quickly find )

2.In accounting for covariates part, you said seed mass is correlated, but the other two are not at the end. I have no clues even I saw the relevant output(plots and summary results). please give me some suggestions and assistance.

  1. and then why we consider to make a multiple regression?(we have suggested the two other factors are not correlated , but we still use them as variables)
roblanf commented 4 months ago

  1. If you see any correlation, you can never know the cause. Let's say you have a correlation between 2 variables, A and B. There are three general causes of such a correlation (assuming it's not chance): A causes changes in B; B causes changes in A; some other variable you haven't measured, call it C, causes changes in both A and B. Of course, all three of these can also be true (and in biology, they usually are all true). I didn't understand the part of your question in brackets.

  2. This was from the single-variable regresssions, and from the multiple regression. We covered this during the workshop itself - check your notes and/or read about multiple regression.

  3. Because a multiple regression may reveal interdependencies in the data that are not apparent from single-variable regressions. In other words, it's possible that variable C does not have a significant correlation with A in a single variable regression, but that it is correlated in the context of a multiple regression with other variables e.g. A ~ B+C+D.