1, 2. How much of a background do you have running time series analysis in linear algebra? The best fit line through your data set is going to be something called the Least Squares Approximation of the the points on the data set. This approximation technique works for n-many dimensions for arbitrarily many data points at countably many times. Consequently, your RSE also depends on this.
http://www.bionicturtle.com/how-to/article/residual-sum-of-squares-and-standard-error-of-regression
3. Percent of variation: Simply your R^2 value, otherwise known as the coefficient of determination.
4. Self Explanatory
5. P-values: This is, given some random variable X with CDF and PDF , find P{X <= a} for arbitrary a along in the graph of the CDF given a
true null-hypothesis. Remember that, to be a CDF, the function defined on the graph of variables used (we'll say x-y) must satisfy the triplet (Omega, F, P), P being a probability measure and F a sigma algebra embedded in the field Omega, with CDF being piecewise right-continuous and have a sum of 1, with PDF being the derivative of the CDF. Omega must be topologically connected with uniform density in this case. Your job in this case is essentially to find the Probability of any arbitrary value. This should be easy for you as a stats person, as I'm assuming you've taken probability. If not, fuck you.
6. easy.
7. I'm not sure about this one, so I can't help you. I've never done one of these before.
