Two questions on Quant

Hi Guys, Hope your preparations for L2 is in full swing. Pls bear with my two simple queries on Quant, 1) For all co-efficients(Intercepts and Independent Variables) they give a Standard Error, how they calculate this Standard Error? 2) For Linear Regression, we arrive at the Slope Co-efficient by doing covar(x,y)/var(x) and then putting mean of X and Y to get Intercept. How they get Intercept and Slope Co-efficient for Time Series data as there is no X and Y, its only one series of data. Pls help me get these basic confusions out of my way. Many thanks in advance.

  1. It’s the standard deviation of coefficients. So for instance, let’s say you do a single variable regression that produces some empirical slope, B. Not all the points (if any) will fall exactly on this slope. The distance between B and each data point is represents error in the regression. This is used to calculate the Standard Error. (More specifically, it’s the sum of distances squared over over N-1.) 2) I don’t understand the question. Maybe someone else…
  1. concur with above 2) X is time, so slope coefficient is Y at t = 0, and the coefficient for X is the slope, so really the average change in Y that results from another unit of time occuring (X=1 to X = 2 for example), standard rise over run from high school, brah. slope coefficient b1 = (change in Y)/(change in X) = (Y2 - Y1)/(X2 - X1) and then you can do your confidence intervals based on the above, which is your b with a hat on top of it; confidence interval b1 = b1

ohai Wrote: ------------------------------------------------------- > 1) It’s the standard deviation of coefficients. So > for instance, let’s say you do a single variable > regression that produces some empirical slope, B. > Not all the points (if any) will fall exactly on > this slope. The distance between B and each data > point is represents error in the regression. This > is used to calculate the Standard Error. (More > specifically, it’s the sum of distances squared > over over N-1.) > > I don’t think this answers his question - I believe this is the answer to what the standard error is for the entire regression, not just the coefficients… It sounds like he’s asking about the standard error for each coefficient…

Thanks guys. But I want the Std Error for each coefficient, not SEE, I understand SEE of the Regression.

verse214 Wrote: ------------------------------------------------------- > ohai Wrote: > -------------------------------------------------- > ----- > > 1) It’s the standard deviation of coefficients. > So > > for instance, let’s say you do a single > variable > > regression that produces some empirical slope, > B. > > Not all the points (if any) will fall exactly > on > > this slope. The distance between B and each > data > > point is represents error in the regression. > This > > is used to calculate the Standard Error. (More > > specifically, it’s the sum of distances squared > > over over N-1.) > > > > > > I don’t think this answers his question - I > believe this is the answer to what the standard > error is for the entire regression, not just the > coefficients… It sounds like he’s asking about > the standard error for each coefficient… The standard error for each coefficient is determined in a different way - it depends on which coefficient you are looking for … For example, for the slope, the standard error would be s=SEE/√[(n-1) sx^2]

quant is the biggest waste of time on this exam

schweser says , std error will be most likely given , however there is one formula for calculating standard error of the forecast (Schweser Pg 154) S^2 f=SEE^2 [1+1/n+((X-Xbar))/((n-1) s^2 x)]

The formula given by Factor hedge is to find out the confidence interval for the estimated value of Dependent Variable for any given value of Independent Variable. This is not what i am looking for, I am searching for the method in which they calculate the Std Error for Intercept and Independent Variable co-efficients. This Std Error we use as the denominator to calculate t statistics for the co-efficients.

Hi Omar, I could not quite follow the formula specified by you, could you pls elaborate? Sorry, if I sound dumb!

sasankm Wrote: ------------------------------------------------------- > Hi Omar, > > I could not quite follow the formula specified by > you, could you pls elaborate? Sorry, if I sound > dumb! it’s actually me the dumb here who copied it from pdf file without editing :slight_smile: Find the following PDF file to understand how to find standard error for each term https://fileshare.aus.edu/pnp7/files/dKgMPwmQc/Chapter_16.pdf Pages 7 - 8 Good luck Dear

Factor hedge Wrote: ------------------------------------------------------- > schweser says , std error will be most likely > given , however there is one formula for > calculating standard error of the forecast > (Schweser Pg 154) > > S^2 f=SEE^2 [1+1/n+((X-Xbar))/((n-1) s^2 x)] This is the formula I think he linked to on pg 8 (middle). This is correct and will be GIVEN on exam. Not sure if that answers your question or not.

pretty confusing thread. Just know all the formulas and how to do the EOCs and you’ll be fine. I didn’t see any questions that asked you to draw a line of best fit for a time series, or asked how the coefficents of a time series were calculated, so I really don’t think you need to know that for the exam; it will be given. You’ll just need to test its significance, or find flaws in the model (serial correlation by looking at the DW value, or heteroskedasticity by doing a BP test).