John Dodson:hi Holly
  Holly Borgie:Hello
  Greg:I can hear.  Hi John.
  John Dodson:ok thanks
  Greg:John, each time I try, I get a negative beta for CSCO.  Omega and alpha
are positive, though.  Have you had success with CSCO this time around?
  Holly Borgie:that one was pesky for me too
  Holly Borgie:I didn't get a negative beta though when I changed my starting
values, but the numbers look kind of odd
  John Dodson:I got valid answers, but they were out of line with the others
for CSCO
  John Dodson:A much higher omega and a lower beta
  Greg:Thanks.  I'll keep trying.
  Scott:could we get the conference call going?
  John Dodson:Scott, I can open it if it is necessary. What do others think
about opening the phone line?
  Greg:Sounds fine. I'll dial in.
  Scott:whats the number again?
  sangya:i'm getting a negative beta for that one too
  John Dodson:Sangya, do you have the little yellow telephone icon in the
corner? roll over that.
  sangya:biz days
  Holly Borgie:John, quick question.  Can I give you a workbook with Bloomberg
links in it, or do you want me to remove them?
  John Dodson:Holly, you should probably sever links. I don't have Bloomberg on
my mac.
  Holly Borgie:ok, sounds good.
  John Dodson:thanks.
  John Dodson:If you are fighting with negative betas, I recommend you
re-formulate your problem in terms of log(alpha) and log(beta).
  John Dodson:The log transofrmation is easy to do and it prevents negative
alphas and betas (although not necessarily negative omegas).
  sangya:i'll try that
  Greg:Here's an off-topic question.  Do you know who our next instructor is,
and if he/she has a website?  The MFM site doesn't ever seem updated with
current instructor website information.
  John Dodson:Greg, it is Blaise Morton. He is a local fixed income arb
manager. I don't think he has a website.
  Greg:Ok. Thanks.
  John Dodson:The third instructor is Arkady Shemyakin, who is a local
academic. He will have a website. I don't know if he will broadcast or record
sessions. I am pretty sure Blaise won't.
  Holly Borgie:confirmed - I emailed him - he won't be recording
  sangya:i keep getting an error that says to increase MaxFunEvals option. does
that mean there is probably something wrong with my code?  any idea?
  John Dodson:Sangya, it probably has to do with your initial guess. Maybe it
is spending too much time looking at negative values.
  sangya:oh ok...maybe the log transformation will help
  sangya:I have to get going, thanks john
  John Dodson:ok, bye Sangya
  yiran 3:John, all gammas here are negative?
  John Dodson:Yiran, I have found that gammas are generally negative. I was
surprised at the non-negative value for GOOG
  yiran 3:me too...
  yiran 3:Is it a lot of work to figure out queation 2?
  John Dodson:Yiran, I don't think so. But you have to be a little careful. You
have to loop over all possible pairs of dates and measure the concordance for
each pair of stocks for each pair of dates.
  yiran 3:thanks!
  yiran 3:Is that to say, when 11.2 is coming, we can check our model to see
whether it is good or not, if nothing unusual happens?
  John Dodson:no, the prices on 11.2 won't tell you much about their risks as
of 10.19.
  yiran 3:So what is question 1 for?
  John Dodson:Yiran, from the point of view of 10.19, the prices on 11.2 are
random and we can only describe them as random variables. Once the market
closes on 11.2, they are no longer random.
  John Dodson:But we still need to work with a model for risk on 10.19, becasue
we have to make investment decisions.
  yiran 3:Okay. Variance just tell us the volatility. So how can I know whether
I should buy or sell. To check the trend of prices?
  John Dodson:Yiran, it will make more sense once we introduce models for
investor preferences next term.
  yiran 3:ok
  John Dodson:Generally you will buy when the risk is small and buy when it is
large.
  John Dodson:sorry... sell when he risk is too large
  yiran 3:make sense
  John Dodson:But "risk" will be evaluated in a portfolio setting with a
particular investor preference.
  Grant:John, I'm having a little trouble getting from the scaled h to the
return forecast. I have h_T+10 and m_T+10 is zero, but what should I use for
Z_T+10 to get Y_T+10?
  John Dodson:Grant, you need something like $h_+E[h_]+E[h_]+\cdots+E[h_]$
  John Dodson:This will be your variance forecast for the log-return
  John Dodson:which you need to convert into a standard deviation forecast for
the price
  Grant:ok, that helps. Thank you.
  Mac:John, is $h_+E[h_]+E[h_]+\cdots+E[h_]$ or $h_T+E[h_]+E[h_]+\cdots+E[h_]$.
Which one is correct for the forecast?
  John Dodson:Mac, the last $h$ from your quasi-MLE objective function was the
forecast for the last residual that you observed. From this, you can forecast
the one-day ahead residual based on the updating equation.
  John Dodson:This is for the return on 10.20. From here you have to use the
expectations, conditional on 10.19, for the conditional variances out to 11.2
  John Dodson:There should be ten terms
  Mac:ok! Thanks.
  John Dodson:Any other questions? I will be here until 9:00 in any case.
  Grant:none for me. Thanks, John. Have a good night.
  Greg:Thanks for your help, John.  Have a good night.
  John Dodson:bye Greg
  Noppanan:Thank you John. Bye