Implementation of a Mortgage Backed Security

(MBS) Pricing Model

S. Zain Hoda and Rui Xiong Kee

May 24, 2007

ORF 535: Computational Finance in C++

Prof. René Carmona
Princeton University
Bendheim Center for Finance

Abstract
We implement a Mortgage Backed Security (MBS) pricing tool. The model
employs a Hull-White single-factor short rate model calibrated to the swaption
volatility matrix. Mortgage rates are determined using a regression on 10 year
treasury yields, using both the original and first-differenced time series as a
check against spurious regression. Our pricing tool finds the price of an MBS
given the OAS or vice versa. Additionally, given the price it can find the Zero
Volatility (ZV) spread. It also determines duration and convexity for hedging
purposes. Regular (non-callable) bonds exhibit a linear relationship between
coupon rate and price. Mortgage Backed Securities exhibit a negative convexity
due the ability of borrowers to prepay. Our pricing tool exhibits the basic
property of negative convexity.1

1 Please note that this document is not as long as it appears; for readability

we used large margins and we have many figures embedded into the document.
The LATEX code we used to produce this document is slightly over 8 pages.

1
1 Introduction
Mortgage Pass-Through Securities, more commonly known as Mortgage Backed
Securities, are bonds that are backed by a pool of mortgages. They are extremely
liquid and are a very important part of the market, and were in the spotlight
recently due to the flurry of recent market activity concerning subprime mort-
gages. The most important differentiating factor of an MBS and a regular bond
is that of prepayment.

In the discussion of the math, we also provide a reference to respective seg-

ment of the program that implements the math-finance topic discussed in a
particular section.

2 Components of the Model

2.1 Interest Rate Model
Class: SimulatedTermStructure
We take as an input today’s instantaneous forward curve. To achieve this we
took eight points on the LIBOR swap curve as the yield curve and fit a smoothed
spline through those points. Points on the instantaneous forward curve are then
given by:
∂
f (t, T ) = y(t, T )(T − t) + y(t, T ) (1)
∂T
Given today’s forward curve we produce simulations of the interest rate over the
time period of the mortgages we are attempting to price. A somewhat simple
way to do this is to use Hull-White’s Extended Vasicek model so that we can fit
the current forward curve perfectly. We used the following form of the Extended
Vasicek.
drt = a(r¯t − rt )dt + σdWt (2)
1 ∂f (t0 , t) σ2  
r¯t = f (t0 , t) + + 2 1 − e−2a(t−t0 ) (3)
a ∂t 2a
It remains to choose parameters a and σ. In an ideal setting, we can find a and
σ by calibrating our model to the market ATM swaption volatility matrix. We
have as input the market implied swaption volatilites. Black’s swaption pricing
formula then gives us the swaption prices:
n
X
P S Black (0) = [Sn (0)N (d1 ) − KN (d2 )] p(0, Ti )τi (4)
i=1

log( SnK(0) ) + σimp

2
T0
d1 = √ (5)
σimp T0
p
d2 = d1 − σimp T0 (6)

2
On the other hand, we can arrive at an estimated swaption price using Monte
Carlo estimation of the expectation, under the risk-neutral measure, of the
discounted payoff of the swaption:
R T0 n
X
EtQ [e− t rs ds
(Sn (T0 ) − K)+ τi P (T0 , Ti )] (7)
i=1

τi is Ti − Ti−1 ; Ti are the LIBOR fixing dates, where P (t, T ) is the price of a
zero coupon bond starting at t and maturing at T . The swap rate Sn (t) is given
by:
1 − P (t, TN )
Rswap (t, T ) = PN (8)
j=1 P (t, Tj )τj

P (t, T ) = e−A(t,T )−B(t,T )rt (9)

−a(T −t)
1−e
B(t, T ) = (10)
a
Z T T
σ2
Z
A(t, T ) = − B 2 (s, T )ds + ar¯t B(s, T )ds (11)
2 t t

Calibration then is the process of finding a and σ to minimize the squared

difference between the market swaption prices and the Monte Carlo estimated
prices:

min [Swaptionmkt − SwaptionM C ]2 (12)

a,σ

In practice we are faced with two constraints. The first is that we only have two
time-homogeneous parameters, a and sigma, with which to potentially match
many swaption prices. The second is that using our small scale implementation
to calibrate to several points may be computationally prohibitive. As such, we
decided to match one point of the volatility matrix, which we choose to be the
5yr5yr ATM2 swaption volatility.

We employ a 2-step procedure to calibrate our model to the 5yr5yr swaption

volatility under our scaled-down approach. In the first step, we use historical
data to estimate “reasonable” values for a and σ. In order to do this, we consider
the regular Vasicek model:

drt = a(r̄ − rt )dt + σdWt (13)

Which we can rewrite as an AR(1):

rt = (1 − ρ)α + ρrt−1 + t , (14)

2 2
where t ∼ N (0, σ∆ ) with σ∆ = σ 2 (1 − exp(−2κ∆))/(2∆), ρ = exp(−κ∆).
Where ∆ is 1/frequency of data. Using 3 month T-bill rates as a proxy for the
25 year swap rate starting 5 years from today

3
short rate, and considering weekly data (∆ = 1/52) from the Federal Reserve
Board from 1982 - 2004, we find the following estimates 3 :
σ : 0.01219293
a : 0.1686036
With these values in mind, we fix the value of a, varying only the σ parameter
in order to match the Monte Carlo price to the market swaption price.

µ = Mean of market implied swaption volatility (5/2006 - 5/2007)

σ = SD of market implied swaption volatility (5/2006 - 5/2007)
From the table above, we see that the choice of a = 0.1 strikes a good balance
between keeping a relatively close to its historical value, while simultaneously
ensuring that the σ we get from calibration is close to historical sigma.

Mean = 15.77, Standard Deviation = 0.9841

2.2 Mortgage Rate Model

Class: MortgageRateGenerator
Since mortgage rates are highly correlated to swap rates, one method for simu-
lating mortgage rates is to take simulated 10-year swap rates and regress these
3 see appendix for S-plus code

4
with the historical mortgage rates. Ordinarily, one would compute the future
swap rates. However, we chose to get the 10-year yield curve from the short
rates to save on computation time. We performed a regression of historical 30-
year fixed mortgage rates against 10-year treasury yields over the period 1994
to 2003.

Residual Standard Error = 0.0024, Multiple R-Square = 0.9179

N = 2501, F-statistic = 27947.97 on 1 and 2499 df, p-value = 0

numeric matrix: 2 rows, 4 columns.

coef [Link] [Link] [Link]
Intercept 0.0289 0.0003 106.7269 0
X 0.7813 0.0047 167.1765 0
There is always the worry that these are two nonstationary time series. To be
more rigorous, we check to see if they are cointegrated variables whose first dif-
ferences are stationary. This check involves performing a unit root test on the
regression residuals. Based on the Augmented Dickey-Fuller test we reject, at
the 5% critical level, the null hypothesis of a unit root with a p-value of 0.02744
and thus find evidence in favor of cointegration.

Another way to avoid the problem of nonstationary variables is to do a re-

gression of their first differences instead.
Residual Standard Error = 0.0001, Multiple R-Square = 0.0367
N = 2500, F-statistic = 95.0948 on 1 and 2498 df, p-value = 0

coef [Link] [Link] [Link]

Intercept 0.0000 0.0000 -2.3217 0.0203
X 0.0346 0.0035 9.7517 0.0000
We see that the relationship between the first order difference is also signif-
icant. We can check how well the coefficients of the regression and the first
difference regression are able to predict the mortgage rates. Using the formulas
Ŷt+1 − Yt = 0 + 0.346(Xt+1 − Xt ) and Ŷt+1 = 0.0289 + 0.7813Xt+1 we find
that our predicted Ŷt + 1 are similar (see graph). As such we conclude that us-
ing the regression coefficients of mortgage rate on 10 year treasury yield is valid.

5
We have provided a C++ implementation of ordinary least squares regression.

2.3 Prepayment Model

Class: cprPathGenerator

Given the mortgage rate, we want to consider when the borrowers will prepay
their loans. We have provided two implementations of a prepayment model
(called A and B in the program). Prepayment model A is modeled after slides
presented by John Naud whereas prepayment model B is taken from a Fannie
Mae paper (Chen 2004) to use as a check. Many factors contribute to when the
prepayments occur. For simplicity we choose a few of these factors and attempt
to model them.
Model A uses the following formula to generate CPR (Constant Prepayment
Rate - the annualized rate of prepayment at each time):

CP R(t) = RI(t)BM (t) + AGE(t)M M (t) (15)

while Model B uses:

CP R(t) = RI(t)AGE(t)M M (t)BM (t) (16)

6
Where RI(t) is the refinancing incentive, BM (t) refers to burnout, AGE(t) is
the seasoning ramp, and M M (t) is the monthly multiplier. Each of these factors
of CPR are explained in individual sections to follow.

2.3.1 Refinancing Incentive

Mortgage borrowers have the option to prepay their loans and take out a new
loan with a lower mortgage rate. If borrowers exercised this option optimally,
we would see that anytime the current mortgage rate falls below the mortgage
rate of the borrower, they would refinance (assuming no transaction costs).

Empirically borrowers do not optimally excercise this option, resulting in an

S-curve.

(Courtesy Naud 2007)

For prepayment model A, we start by observing that the S-curve graph of CPR
to incentive can be replicated by a normal cumulative distribution function

7
(CDF), suitably shifted and rescaled. Our refinancing function thus employs
the normal CDF:

RI(t) = Φ((incentive − shift)rescale) (17)

Using appropriate values for shift and rescale.

Such an approach not only ensures the replication of empirically observed behav-
ior, it also has theoretical underpinnings. For instance, Kariya and Kobayashi
(2000) suggest a model for the exit time of the kth mortgagor as:

τk = min{j : W AC − Rj ≥ cj (k)} (18)

Where cj (k) denotes the “incentive threshold” of the kth mortgagor at time j,
W AC refers to the weighted average coupon at the time of the mortgage pool’s
issue (i.e. the mortgage rate we start with), and Rj refers to the mortgage
rate prevailing at time j. Note that if all mortgagors were optimal exercisers
of their option to refinance, cj (k) should be 0. In the simplest model with no
information on the kth mortgagor is available, Kariya and Kobayashi suggest
that we can explain the heterogeneous response of mortgagors to refinancing
incentive by having incentive thresholds cj (k) that are normally distributed in
the mortgage pool. This is equivalent to having the refinancing CPR function
in the form of a normal CDF.

Model A uses the cumulative distribution function for a Gaussian random vari-
able to produce the so-called “S-curve” plus a seasoning ramp, whereas Model
B uses the following formula:

RI(t) = 0.28 + 0.14 tan−1 (−8.571 + 430(W AC − r10,t−1 )) (19)

8
2.3.2 Seasoning Ramp
People do not tend to move (and hence prepay) within the first few months of
buying a new house. The prepayment rate increases as the loan age increases
up to a certain point.
The PSA model is one of several models used to calculate and man-
age prepayment risk. The PSA model acknowledges that prepay-
ment assumptions will change during the life of the obligation and
affect the yield of the security. The model assumes a gradual rise in
prepayments, which peaks after 30 months. The standard increase
amount is 0.2%, so an indicator of 100% implies that the rate will
increase monthly by 0.2% (the standard increase), whereas an indi-
cator of 0% implies no monthly changes of the prepayment rate.4

 
t
AGE(t) = min 1, (20)
30
4 [Link]

9
2.3.3 Monthly Multiplier
M M (t) is [0.7, 0.65, 0.85, 0.95, 1.0, 1.2, 1.1, 1.35, 1.15, 1.1, 1.0, 0.95] respective to
the month due to the fact that turnover in the housing market varies by month.5

2.3.4 Burnout Multiplier

Burnout refers to the empirical slowdown of prepayments due to mortgage-rate
changes following periods of intensive refinancings. The people who are most
likely to respond to incentives to refinance do so early, leaving the pool with
borrowers who are less likely to refinance. 6
ˆ
Bt−1
BM (t) = 0.3 + 0.7 (21)
B̂0
Where Bˆn is defined in equation 31
5 Andrew Davidson & Co Research Report
6 Kalotay Yang Fabozzi 2004

10
2.3.5 Other Approaches to Modeling Prepayment
Kalotay, Yang and Fabozzi (KYF) treat the mortgagor’s right to refinance as
an American call option on mortgage rate with the weighted average coupon as
its strike. At each time, the mortgagor compares the value of exercise to the
value of waiting. In order to account for sub-optimal exercise of the option,
KYF create a distribution of strike rates across the population of mortgagors.
In other words, mortgagors who are slow to refinance (i.e. wait till the option is
well into the money before exercising) do so because they believe that the strike
is higher than it actually is, while mortgagors who refinance too soon believe
that the strike is lower than it is.

A much more basic attempt to model refinancing behavior is described in Lon-

don (2005). Here, refinancing CPR is fixed between certain bounds:


0 if R0 − Rj > k4

b4 if k3 < R0 − Rj ≤ k4





b if k < R − R ≤ k
3 2 0 j 3
RI(j) = (22)


b 2 if k1 < R 0 − R j ≤ k2
b1 if k0 < R0 − Rj ≤ k1




1 if R0 − Rj ≤ k0


Where k4 > k3 > k2 > k1 > k0 and 0 < b4 < b3 < b2 < b1 < b0 < 1 are
positive constants. R0 and Rj have the same meaning as in equation 22. In
this approach, we are simply using a step function to approximate the S-curve
of CPR against incentive.

11
The method described in London’s book is too simplistic for our purposes. How-
ever, there is a strong desire within academia to employ an options framework to
analyze refinancing behavior, and in this respect, the methodology suggested by
Kalotay, Yang and Fabozzi is appealing. However, we believe that our approach
circumvents the modeling complexity of American options, while retaining the
spirit of the approach - using a distribution of incentive thresholds to explain
the suboptimal option exercise behavior of mortgagors.

2.4 Cashflow Model

Class: CFgenerator

Below is a table of several CPR scenatios and their cashflows. As a sanity check,
we note that our cashflow generator matches the output in Naud (2007).

The cashflow model is just a straightforward implementation of the formulas

below.

12
Zero Prepayments:
CFn = SPn + In (23)
dN −n+1
SPn = g (24)
1 − dN
1 − dN −n+1
In = c (25)
1 − dN
1 − dN −n
Bn = (26)
1 − dN
Nonzero Prepayments:
CFˆn = ˆ n + P ˆPn + Iˆn
SP (27)
ˆn =
SP Qn−1 SPn (28)
ˆ
P Pn = ˆ − SP
(Bn−1 ˆ n )SM Mn (29)
Iˆn = Qn−1 In (30)
Bˆn = ˆ − SP
Bn−1 ˆ n − P ˆPn = Qn Bn (31)
Where g is the gross monthly coupon on underlying loans; c is the net monthly
coupon received by MBS holder, N is the original loan maturity in months; n is
the loan age in months; Bn is the remaining principal balance per unit of original
balance after nth monthly payment assuming zero prepayments; d = 1/(1 + g);
(1 − SM M ) = (1 − CP R)1/12 ; Fn is the pool factor (actual remaining balance
per dollar of original principal); Qn = Fn /Bn is the fraction of the pool that
has not yet prepaid; Qn = Qn−1 (1 − SM M ), Q0 = 1.

3 Combining the Components

Class: MBS
This class starts the chain of computations and simulations given the necessary
input parameters to the model.

3.1 Computing the Option Adjusted Spread or Price

Method: MBS->OAspread
MBS->getPrice
Given cashflows CFj , asset pricing would tell us that we would just have to
discount these cash flows under the risk neutral measure to produce the price.
 
XN R tj
pt = EtQ  e− t rs ds CFj  (32)
j=1
 
XN R tj
= EtQ  e− t f (s,s)ds CFj  (33)
j=1

13
However, we could only do this if we knew when people prepayed. Unfortunately
prepayment is not determinstic nor is it excercised optimally so we have to
simulate this as well and there will be a certain error in the price. To account
for this error, we add the Option Adjusted Spread.
 
XN R tj
pt (λ) = EtQ  e− t f (s,s)ds−λ(tj −t) CFj  (34)
j=1

Given either the price or the OAS, using this formula we can compute the other
value.

3.1.1 Computing the Zero Volatility Spread

Method: MBS->ZVspread

We built the apparatus to compute the OAS, which is a more difficult problem.
To compute the ZV spread, we go through the same process as for OAS except
instead of using simulated interest rates we use today’s deterministic forward
rates and use λ0 instead of λ.
N
X R tj
pt (λ) = e− t f (t,s)ds−λ0 (tj −t)
CFj (35)
j=1

3.2 Comparing Our Spreads with Bloomberg

We compare our ZV spread and our OAS with Bloomberg. For both, we have
spread values that are around the same order as Bloomberg. Our ZV spread
more closely matches Bloomberg values because only the cashflows are pre-
dicted; the discount rates used are today’s forward curve.

14
We would expect the OAS to diverge more from Bloomberg because of the added
prediction of interest rates, which is again what we see.

3.3 Duration and Convexity

Method: MBS->OAduration
MBS->OAconvexity

The mathematically correct method to compute duration and convexity is to

compute the first and second derivatives. However, a common numerical al-
ternative is effective duration and effective convexity. We use ∆λ of 50 basis
points.

P (λ − ∆λ) − P (λ + ∆λ)
Def f = (36)
2P (λ)∆λ
P (λ + ∆λ) + P (λ − ∆λ) − 2P (λ)
Cef f = (37)
P (λ)(∆λ)2

The following graph shows the price versus the coupon rate for a normal (non-
defaultable) bond. We see that as the coupon rate increases, the price of the
bond increases linearly because the cash flows from the bond increase in a linear
manner with coupon rate. We provide this as a contrast to the behavior of the
Mortgage Backed Security. The parameters used are somewhat arbitrary; we
only mean to show the linear relationship.

15
The following graph shows the price versus coupon rate for an MBS generated
by our model. Immediately we notice that it has a negative convexity. This
negative convexity is due the prepayment property of MBS. Up to a certain
point, the MBS acts like a Treasury bond. However, once the coupon rate (and
hence the correponding mortgage rate) becomes higher, the borrowers have more
incentive and opportunity to refinance. The refinancing causes prepayments,
making the bond less valuable than it would otherwise be.

3.4 Hedging
Now that we have derived a price for MBS from our model, the question remains
as to how to hedge our position in MBS.

16
3.4.1 First approach - Duration and Convexity Hedging
Unfortunately, this is an unusually difficult endeavour. Even a simple strategy
involving duration and convexity hedging faces the problem of an appropriate
definition of duration and convexity7 . However, if we follow the current conven-
tion of OAS duration and convexity, hedging becomes a matter of holding the
portfolio P + q1 H1 + q2 H2 where P is the position in MBS, Hi is the price of a
hedging instrument and qi is the quantity of the hedging instrument. We solve
for qi using:

P 0 + q1 (H1 )0 + q2 (H2 )0 = 0 (38)

P 00 + q1 (H1 )00 + q2 (H2 )00 = 0 (39)

Where P 0 refers to the derivative of the price of MBS with respect to yield, i.e.
duration, and P 00 refers to convexity. Typical hedging instruments are Treasury
note futures, interest-rate swaps and swaptions, as well as MBS with different
coupon rates.

3.4.2 Second Approach - Regression-Based Hedging

A solution to the problem of model-dependency is an “empirical” hedging strat-
egy, where the hedge ratio is estimated from a simple regression:

∆St = α + β∆Ft + t (40)

(41)

Where ∆ is the first-difference operation, St is the price (or the logarithm of the
price) of MBS at time t, Ft is the price (or logarithm) of the hedging instrument
at t, t is the i.i.d. random error process. The hedge ratio is then the estimated
β.

3.4.3 Third approach - Incorporating the dynamic link between MBS

and Treasurys
More recently, Bhattacharya, Sekhar and Fabozzi (2006; BSF hereafter) suggest
that hedging strategies should take into account the dynamic link between the
MBS and treasury markets. For instance, they posit that the hedging activity
of MBS investors in the swap and treasury markets increases the volatility of
interest rates, which in turn leads to higher hedging requirements by holders
of large MBS portfolios. Given the size of the MBS market, they quote the
popular adage that the “mortgage tail is wagging the Treasury dog,” suggesting
that the dynamic link is indeed an important factor.

BSF use a cointegration model to capture the co-dependence between MBS

and treasurys in terms of a long-run equilibrium relationship in their prices.
Additionally, observing that volatility in the two markets is persistent, they
7 see Fabozzi 2004 for discussion

17
model volatility as a GARCH process.
They pick five securities to include in their cointegration model, and their long
run relationship is given as:

Ptt−note = λ0 + λ1 Pttba5.5 + λ2 Pttba6.0 + λ3 Pttba6.5 + λ4 Pttba7.0 + Corrt (42)

The corresponding error correction model for the log prices of the five securities
is:
∆PtX = αtX + βtX (Corrt−1 ) + γtX (M vart−1 ) + εX t (43)
Where X takes the values {tba5.5, tba6.0, tba6.5, tba7.0, t − note} and the error
terms of equation 43 are assumed to follow a multivariate normal distribution,
and are given by εt = [εX Xn
t · · · εt ], with the GARCH structure given by
1

Ht = A0 A + B 0 εt−1 ε0t−1 B + C 0 Ht−1 C (44)

where Ht is the variance-covariance matrix for the variance equations of the

GARCH processes for the five securities at time t, A is the diagonal matrix of
intercepts and B and C are the diagonal matrices of slope coefficients of the
variance equations. Hedge ratios are calculated by
σtba5.5,tba6.0
wtba6.0 = − 2 (45)
σtba5.5
2
where σtba5.5,tba6.0 and σtba5.5 are part of the variance-covariance matrix gen-
erated by the model. The principal improvement of the BSF cointegration
GARCH model is that it reflects two market realities - that the causal rela-
tionship between MBS and treasury prices is two-way, and that volatility is not
constant across time - that the simple regression model does not.

18
4 Implementation

We employ inheritance in the following manner so that we can compute either

the Zero Volatility Spread or the Option Adjusted Spread without altering too
much code.

19
5 Possible Extensions to Current Implementa-
tion
There have been several papers suggesting that house prices are a significant
factor in determining prepayment rates. For instance, declines in house prices
reduce mortgagors’ credit quality, making it harder for them to refinance or take
out new loans, thus reducing the likelihood of them moving and prepaying their
existing loan (Downing, Stanton and Wallace 2005). On the other hand, rising
house prices may encourage mortgagors to refinance - taking advantage of their
more valuable collateral to increase the size of their loan (Lee and Pace 2006).
Some papers have thus sought to model

HI(t) = (Ht − H0 ) (46)

where HI is the incentive, Ht is the current regional house price and H0 is the
regional house price at the start of mortgage) as another factor in prepayment.

In general, incorporating pool-specific information augments the precision of

the prepayment model. Details such as average credit quality, possibly proxied
by median income, are often incorporated in prepayment models. Where a large
number of factors are included, dimension-reduction techniques (e.g. PCA) are
employed.

6 Conclusion
After programming the various components of the MBS pricing model in C++
we were able to generate reasonable ZV and OAS values. We also satisfy the
basic check of negative convexity. Furthermote, we have ensured that our pricing
tool exhibits sound fundamentals - a short rate model calibrated to market
swap and swaption prices, and a prepayment model that is supported by both
theory and historical data. Finally, we discussed several extensions to our simple
implementation in terms of short rate modeling, prepayment modeling, and
hedging. These are possible areas for future improvement.

20
References and Data Sources
6.1 References
John D. Naud, ”Valuation and Hedging of Mortgage Backed Securities.” April
18, 2007
Jian Chen, ”Simulation-Based Pricing of Mortgage Backed Securities.” 2004
[Link]
Justin London, Modeling derivatives in C++. 2005
John Hull, Alan White. ”Pricing Interest-Rate-Derivative Securities.” 1990
Takeaki Kariya, Masaaki Kobayashi. ”Pricing Mortgage-Backed Securities (MBS).”
2000
Andrew Kalotay, Deane Yang, Frank J. Fabozzi. ”An Option-Theoretic Pre-
payment Model for Mortgages and Mortgage-Backed Securities.” 2004
Min-Long Lee, R. Kelley Pace. ”Local housing prices and mortgage refinancing
in US cities.” 2006
Chris Downing, Richard Stanton, Nancy Wallace. ”An Empirical Test of a Two-
Factor Mortgage Valuation Model: How Much do House Prices Matter?” 2005

6.2 Data Sources

Mortgage Rates:
[Link]
Treasury Yields:
[Link]
debt-management/interest-rate/yield_historical_huge.shtml

21
Appendix: S-Plus Code for Historical Estimation
#a
dt<-1/52
onemotime<-matrix(1,1160,1)
for(i in 1:1160){
onemotime[i] = threemo[i,2]
}
onemodp<-onemotime/100 #change to decimal points from percentage
alpha<-mean(onemodp)
alphavect<-matrix(alpha,1159,1)

rtvect<-onemodp[2:1160]
rt1vect<-onemodp[1:1159]
rhonumerator<-sum((rtvect-alphavect)*(rt1vect-alphavect))
rhodenominator<-sum((rt1vect-alphavect)^2)
rho<-rhonumerator/rhodenominator

rthat<-(1-rho)*alphavect + rho*rt1vect
et<-rtvect - rthat
sigdel<-sum(et^2)/length(et)

kappa<-(-log(rho)/dt)
sig<-sqrt(2*kappa*sigdel/(1-exp(-2*kappa*dt)))

22