WEBVTT
Kind: captions
Language: en
00:00:01.540 --> 00:00:09.500
Now we are moving into the one type of renewal
process that is called alternating renewal
00:00:09.500 --> 00:00:19.090
process. Let Xi's are iid random variable
which constitute on times and Yi's are iid
00:00:19.090 --> 00:00:26.550
random variable which constitutes off times.
Assume that mean is the existent it is finite
00:00:26.550 --> 00:00:37.870
and the X plus Y has the distribution F. Suppose
that the renewal occurs at the end of every
00:00:37.870 --> 00:00:49.050
Xi's whereas no arrivals at the end of every
Yi's. Assume that Xi's and Yi's are independent
00:00:49.050 --> 00:01:01.850
random variables also. Then the Xi plus Yi
are called the alternative renewal process.
00:01:01.850 --> 00:01:12.260
See the example for this situation. A machine
works for the time X1 and then breaks down
00:01:12.260 --> 00:01:19.240
and has to be repaired which takes a time
Y1. Then works for the time X2 then it is
00:01:19.240 --> 00:01:30.070
down for a time Y2. That's a second repair
time and so on. So Xi's are nothing but the
00:01:30.070 --> 00:01:39.820
machine works and the Yi's are nothing but
the repair time. If you suppose that the machine
00:01:39.820 --> 00:01:48.880
is good as new after each repair then this
constitute alternative renewal process. So
00:01:48.880 --> 00:01:58.960
this is the example of alternative renewal
process with the proper assumptions.
00:01:58.960 --> 00:02:08.769
Now we are moving into the next renewal processes
that is delayed renewal process. It is not
00:02:08.769 --> 00:02:14.549
always reasonable to insist that the first
renewal occurs at time S naught that is equal
00:02:14.549 --> 00:02:22.560
to 0 in the origin itself, at the time 0 itself.
For instant in applications where the occurrence
00:02:22.560 --> 00:02:31.519
are not or at times when a component of the
system must be replaced. One might well be
00:02:31.519 --> 00:02:41.750
interested in situations where there is already
a working component in place at time 0. For
00:02:41.750 --> 00:02:49.279
this reason we defined a delayed renewal process
to be a sequence S naught, S1, S2 and so on
00:02:49.279 --> 00:02:57.920
where Sn is nothing but S naught plus the
summation of first n Xi's and the inter arrival
00:02:57.920 --> 00:03:07.889
times Xi's are positive and iid random variables
as in the ordinary renewal process. And the
00:03:07.889 --> 00:03:13.689
initial delay S naught which is great or equal
to 0 that is independent of inter arrival
00:03:13.689 --> 00:03:23.299
times Xi's.
Notice that the distribution of initial delay
00:03:23.299 --> 00:03:30.760
random variable X naught is not recurred to
be the same as that of the inter arrival time
00:03:30.760 --> 00:03:38.809
random variables Xi's. Hence a delayed renewal
process is a renewal process in which the
00:03:38.809 --> 00:03:46.709
first arrival time X1 is independently and
is allowed to have a different distribution
00:03:46.709 --> 00:03:55.109
that is F1. The distribution of all remaining
iid random variables that distribution is
00:03:55.109 --> 00:04:06.809
F. So the F1 is different from F. The X1 is
equal to t1 that is nothing but that delay
00:04:06.809 --> 00:04:14.730
and there is no such delay then X1 is also
distributed in the same way so the distribution
00:04:14.730 --> 00:04:30.320
of X1 is F as usual. Then the renewal process
is said to be a non delayed version.
00:04:30.320 --> 00:04:40.780
Now we are discussing the central limit theorem
on the renewal process. As n tends to infinity
00:04:40.780 --> 00:04:47.650
the random variable S suffix n that is nothing
but the nth time renewal minus n times mu
00:04:47.650 --> 00:04:57.260
divided by Sigma times square root of n will
be normal distribution with the mean 0 and
00:04:57.260 --> 00:05:05.900
variance 1. So this convergence takes place
in a distribution. So this is a weak distribution,
00:05:05.900 --> 00:05:11.530
weak convergence.
So as n tends to infinity the random variable
00:05:11.530 --> 00:05:20.800
Sn and the mean of Sn is n times mu and the
variance of Sn is Sigma square n and the random
00:05:20.800 --> 00:05:32.140
variable minus the mean divided by the standard
deviation is normal distributed with the mean
00:05:32.140 --> 00:05:42.210
0 and variance 1 as n tends to infinity as
n tends to infinity the n of t the counting
00:05:42.210 --> 00:05:50.060
process it's a renewal process becomes a normal
distribution in the mean t divided by mu and
00:05:50.060 --> 00:05:58.670
the variance is Sigma square t divided by
mu cube where mu is the mean of inter arrival
00:05:58.670 --> 00:06:09.940
time as a t tends to infinity the random variable
n of t minus t by mu divided by Sigma times
00:06:09.940 --> 00:06:17.280
square root of t by mu cube will be normal
distribution with the mean 0 and the variance
00:06:17.280 --> 00:06:33.240
1. So this is also a convergence in distribution.
Now we are going to discuss the long-run properties
00:06:33.240 --> 00:06:40.960
of renewal process. There are two types of
long run properties. One is to obtain the
00:06:40.960 --> 00:06:49.660
long-run average of the quantity of interest.
The other one is to obtain the pointwise limit.
00:06:49.660 --> 00:06:56.520
For example the long-run average of age that
is limit t tends to infinity the integration
00:06:56.520 --> 00:07:06.139
0 to t A of s, A of s is the age divided by
t. while the pointwise limit of the expected
00:07:06.139 --> 00:07:17.250
age that is the limit t tends to infinity
expectation of A of t.
00:07:17.250 --> 00:07:24.830
One can study the average number of renewals
per unit time in the long run. It is called
00:07:24.830 --> 00:07:34.060
the long run renewal rate. For a renewal process
having a distribution F for the inter arrival
00:07:34.060 --> 00:07:40.930
times the long run renewal rate is nothing
but limit t tends to infinity n of t divided
00:07:40.930 --> 00:07:50.430
by t. That will be 1 divided by mu with the
probability 1 where mu is nothing but the
00:07:50.430 --> 00:08:05.800
mean of inter arrival time.
Since S suffix n of t plus last renewal time
00:08:05.800 --> 00:08:15.850
prior to t and S suffix n of t plus 1 is the
first renewal time of t we know the relation
00:08:15.850 --> 00:08:25.870
of S of n of t with the S of n of t plus 1
and that lies, the t lies between those two
00:08:25.870 --> 00:08:37.329
renewal times.
You can divide by t you can divide by n of
00:08:37.329 --> 00:08:44.529
t. Therefore S of n of t divided by n of t
less than or equal to t divided by n of t
00:08:44.529 --> 00:08:52.990
that is less than or equal to S of n plus
t plus 1 divided by n of t. Now we can evaluate
00:08:52.990 --> 00:09:01.600
the first one the S of n of t divided by n
of t limit t tends to infinity. That is nothing
00:09:01.600 --> 00:09:09.699
but the numerator is nothing but the summation
of Xi's t n of t denominator is n of t and
00:09:09.699 --> 00:09:18.550
that is nothing but expectation of X in a
long run - in the long run the summation of
00:09:18.550 --> 00:09:26.269
X1, X2 till X of n t divided by n of T that
is nothing but the expectation of X that is
00:09:26.269 --> 00:09:33.569
same as the mu with the probability 1. And
similarly one can evaluate the last value
00:09:33.569 --> 00:09:42.360
that is S of n of t plus 1 divided by n of
t that is also will be mu with the probability
00:09:42.360 --> 00:09:50.120
1. Since t lies between S of n of t and S
of n of t plus 1 and the throughout divided
00:09:50.120 --> 00:09:58.279
by n of t in all three therefore as limit
t tends to infinity n of t divided by t will
00:09:58.279 --> 00:10:00.779
be 1 divided by mu with the probability 1.