How to find a 90% or 95% conf Interval , and a suitable length L ( ie number of measures in sample space) to narrow down the confidence interval to 1....

in a bi-variate normal distribution independent forming a linear regression model ? One Linear model in R as example is Give Below.

Expecting answer

X% conf Interval error on estimate as +/- 'e'

If the Sample size if of 'L' for 90% conf Interval , we can have |'e'| as 1.3 units or less.

Note : Substantiate with explanation