query 26

#$&*

course Mth 173

7/16 8:00 pm

026. `query 26

*********************************************

Question: `qQuery Theory 3 Problem 6 (was problem 6 page 186) integral from 1 to 4 of 1 / `sqrt(1+x^2).

How many subintervals did you use, and how are you sure that the lower and upper sums differ by less than 0.1? What is the number of intervals required to ensure a discrepancy of less than .05?

YYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYY

Your solution:

I am not sure how to do this, I know that you will need to use the lower and upper hand sums.

Here’s a guess:

difference between upper and lower sums = 1|f(b) - f(a) | * delta t

So: |f(4) - f(1) * ((4-1)/ 2)

|.243 - .707| * (3/2) = .696

For n = 50: difference = .028

For n = 10: difference = .1392

Trying n= 15: difference = .0928

Trying to get closer to 0.1: n= 14: .09943

But for the second part: number of intervals required to ensure a discrepancy of less than .05?

I guessed and used this way:

difference between upper and lower sums = |f(b) - f(a) | * delta t

delta t = (b-a) / n

since we want the difference to be .05:

.05 = |f(4) - f(a)| * (4-1) / n

.05 n = |.243 - .707| * (3)

n = about 27.84

confidence rating #$&*: 0

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

.............................................

Given Solution:

`aSTUDENT RESPONSE: Sums given. I would estimate that a n-value of 20 would yield a value of .05 of the true value. My reasoning is that when n = 10 there is a difference between the left and right hand sums of .139. And, when n = 15 there is a .093 difference in the left and right sums: .139 - .093 = .046. So, I predict that with another increase of n, by a value of 5 we will gain another .05 (or so) of accuracy. Therfore, my estimate is n = 20 (Though I didn't calculate it)

INSTRUCTOR CRITIQUE:

Good thinking, but since you are dividing by the number of intervals the situation isn't that linear.

The max error is | f(b) - f(a) | * `dx. Since f(1) = 1/`sqrt(2) = .707 and f(4) = 1 / `sqrt(17) = .245, approx., we have max error

| .245 - .707 | * `dx

Setting this equal to 0.1 we get

|.245 - .707| * `dx < 0.1 so `dx < .1 / .462; `dx < = .2 would do.

Since the interval has length 4-1 = 3, then `dx < = .2 implies n > = 3 / .2 = 15.

This agrees with your result.

To get the discrepancy lower than .05 we use the same means, this time obtaining equation

|.245 - .707| * `dx < 0.05, satisfied for `dx < .05 / .462. We could get away with `dx = .11, but .10 is nicer and divides evenly into interval length. If `dx = .10 we get n = 3 / .10 = 30.

** Note that n = 20 won't assure us of a discrepancy of < .05. **

ALTERNATIVE SOLUTION:

** You need to get a good picture of this situation in your head; of course that should probably start on paper.

Sketch a graph of this function from x=1 to x=4 and subdivide into 3 intervals.

Sketch the lower rectangles of the 3-interval Riemann Sum, then the upper rectangles.

Shade in the small rectangles that represent the differences between the upper and lower rectangles.

Now imagine moving each of these rectangles straight to the right, until they're at the right-hand side of your figure, with one on top of the other.

The three rectangles now stack into a single rectangle whose altitude is f(b) - f(a) and whose width is 1, the width of a rectangle.

The area of this stack is the difference between the upper and lower sum, and its area is [ f(b) - f(a) ] * `dt, where `dt is the width of a single rectangle.

Now if you made the `dt smaller, say, .1, this would make all the little rectangles thinner; and when you stacked them the stack would be just as high as before, but skinnier. Its area would be area = diff between upper and lower=(f(b)-f(a))'dt. Since `dt is smaller so is the area, and so therefore is the area.

You found that if `dt = .22, you can reduce the difference to .1.

If `dt is .22 or less, then, the difference will be no greater than .1.

On an interval of length 4 - 1 = 3 this implies that n > 3 / .22 = 13.7 or so; any n >= 14 will do it. **

&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&

Self-critique (if necessary):

I am surprised that I pretty much got this right!

------------------------------------------------

Self-critique Rating: OK

*********************************************

Question: `qQuery Theory 3 Problem 20 (was problem 20 p. 187) prove for continuous f, c in [a,b] that int from a to b = int from a to c + int from c to b

Explain how you have proved the stated result.

YYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYY

Your solution:

This is true since the function on the interval of a to b is continuous then every x value is present on that interval which would include the value of c

From a to c then c to b would just result in a to b, cutting out the middle value

confidence rating #$&*:

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

.............................................

Given Solution:

`a** To understand what this problem is talking about, you can think first about areas. The area under a curve from a to b is equal to its area from a to c, plus its area from c to b. That is, you can split the region under the curves at any x value between a and b and the total area will be the sum of the areas.

The strategy shown in the text is to show that if you find a lower sum from a to c, and a lower sum from c to b, that when you add them you have a lower sum from a to b. Why should this be so?

Then you show that for any lower sum on the interval [a, b], you can find lower sums on [a, c] and on [c, b] whose total is greater. This isn't hard to show because when you refine lower sums by splitting the interval up, the lower sums tend to get larger. So if a = x0 < x1 < ... < xn = b is a partition of [a,b], how can you refine this partition to get partitions of [a, c] and [c, b] for which the lower sum is no less than the lower sum you get over [a, b]? Note that it is not necessarily the case that c is equal to one of the numbers x0, x1, ..., xn, so you can't assume that it is; hence there is at least some work to do here. See if you can do it.

What you show in the first two parts can be used to show that the least upper bound of lower sums on [a, b] is equal to the sum of the least upper bound of lower sums on [a, c] and on [c, b]. Since the least upper bound of lower sums of a continuous function is equal to the integral, this gives the desired result. See if you can put the proof together. **

STUDENT COMMENT

I was under the assumption that the lowest sum between a and c was the lowest lower bound...

INSTRUCTOR RESPONSE

The lowest sum is a lower bound, but so is any number less than this sum.

The lowest sum doesn't exceed the least upper bound.

A lower bound is any quantity which is less than the value you are seeking. There is no limit to how low a lower bound can go. There's a limit to how high it can go--it can't exceed the sought value.

There's also a limit to how low an upper bound can go (the 'least upper bound'), but no limit to how high it can go (no such thing as a 'greatest upper bound').

&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&

Self-critique (if necessary):

I think what I was saying is true, but I didn’t talk about the areas. The explanation about the areas helped me to visualize how to approach the problem.

------------------------------------------------

Self-critique Rating: 3

*********************************************

Question: `qQuery Add comments on any surprises or insights you experienced as a result of this assignment.

YYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYY

Your solution:

I already knew that integrals were kind of hard for me and this is still true, but I am beginning to get a better understanding

"

Self-critique (if necessary):

------------------------------------------------

Self-critique rating:

*********************************************

Question: `qQuery Add comments on any surprises or insights you experienced as a result of this assignment.

YYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYY

Your solution:

I already knew that integrals were kind of hard for me and this is still true, but I am beginning to get a better understanding

"

Self-critique (if necessary):

------------------------------------------------

Self-critique rating:

#*&!

&#Good responses. Let me know if you have questions. &#