Republic of Mathematics blog

Growth rates of simple integer recurrences

Posted by: Gary Ernest Davis on: June 5, 2012

The sequence of integers A(n) defined by A(1)=2 \textrm{ and } A(n)= \lfloor(3/2)A(n-1)\rfloor – where \lfloor x\rfloor , the floor of x , is the greatest integer \leq x – begins 2, 3, 4, 6, 9, 13, 19, 28, 42, 63, 94, 141, 211, 316, 474, 711, 1066, 1599, 2398, 3597, 5395, 8092, 12138, 18207, 27310, 40965, 61447, 92170, 138255, 207382, 311073 and grows rapidly.

One can numerically estimate the growth rate of this sequence by first observing that the sequence of positive numbers \frac{A(n)}{(3/2)^n} is decreasing. This follows from the inequality A(n) =\lfloor(3/2)A(n-1)\rfloor \leq (3/2)A(n-1) upon dividing by (3/2)^n, to get \frac{A(n)}{(3/2)^n}\leq \frac{A(n-1)}{(3/2)^{n-1}}.

This means the sequence \frac{A(n)}{(3/2)^n} has a limit. In fact this sequence converges fairly rapidly to the limit:

n A(n)/(3/2)n n A(n)/(3/2)n
1 1.33333333333 34 1.08151439525
2 1.33333333333 35 1.08151405187
3 1.18518518519 36 1.08151382295
4 1.18518518519 37 1.08151382295
5 1.18518518519 38 1.08151382295
6 1.14128943759 39 1.08151375512
7 1.11202560585 40 1.08151375512
8 1.09251638470 41 1.08151372497
9 1.09251638470 42 1.08151370487
10 1.09251638470 43 1.08151369148
11 1.08673587473 44 1.08151368254
12 1.08673587473 45 1.08151367659
13 1.08416675918 46 1.08151367262
14 1.08245401549 47 1.08151366997
15 1.08245401549 48 1.08151366997
16 1.08245401549 49 1.08151366880
17 1.08194653587 50 1.08151366880
18 1.08194653587 51 1.08151366880
19 1.08172098938 52 1.08151366880
20 1.08172098938 53 1.08151366880
21 1.08162074649 54 1.08151366880
22 1.08155391790 55 1.08151366869
23 1.08155391790 56 1.08151366862
24 1.08155391790 57 1.08151366862
25 1.08153411684 58 1.08151366862
26 1.08153411684 59 1.08151366860
27 1.08152531636 60 1.08151366860
28 1.08151944938 61 1.08151366859
29 1.08151944938 62 1.08151366859
30 1.08151684183 63 1.08151366859
31 1.08151684183 64 1.08151366859
32 1.08151568292 65 1.08151366859
33 1.08151491032 66 1.08151366859

x

So we see that A(n)\approx 1.08151366859(3/2)^n for large n.

What if we replace the constant 3/2 by a parameter r ?

How does the sequence A(n) defined by A(1)=2 \textrm{ and } A(n)= \lfloor rA(n-1)\rfloor grow?

As before we see there is a constant K(r) such that A(n)/r^n \to K(r) \textrm{ as } r \to \infty

An obvious question is: how does K(r) \textrm{ vary with } r ?

A moment’s thought shows that the answer is only interesting for r \geq 3/2

The following Mathematica® code computes K(r) to a decent approximation:

A[1, r_] = 2;
A[n_, r_] := A[n, r] = Floor[r*A[n – 1, r]];
T = Table[{r, Table[N[A[n, r]/r^n], {n, 1, 100}][[-1]]}, {r, 1.5, 5, 0.001}];
ListPlot[T, Joined -> True]

and the resulting plot of K(r) \textrm{ versus } r looks as follows:

A blow-up of the portion of the plot for 1.5 \leq r \leq 2.5 looks as follows:

The Centers for Disease Control and Prevention website has a very interesting series of graphics depicting adult obesity rates in the United States by State and by year from 1985 through 2010.

Obesity for this purpose is defined as a body mass index (BMI) of 30 or above.

The changes year by year are dramatic.

The percentages of adults in the various states that are obese by this definition are shown for 2010 in the graphic below:

The States, buy sorted by percentage of obese adults from largest to smallest, patient is shown in the table below, together i the z-scores for those percentages (for this purpose the District of Columbia is counted as a State)

STATE % PEOPLE WITH BMI >30 Z-SCORE
Mississippi 34 2.17
West Virginia 32.5 1.70
Alabama 32.2 1.61
South Carolina 31.5 1.40
Kentucky 31.3 1.33
Texas 31 1.24
Louisiana 31 1.24
Michigan 30.9 1.21
Tennessee 30.8 1.18
Missouri 30.5 1.09
Oklahoma 30.4 1.05
Arkansas 30.1 0.96
Indiana 29.6 0.81
Georgia 29.6 0.81
Kansas 29.4 0.75
Ohio 29.2 0.68
Pennsylvania 28.6 0.50
Iowa 28.4 0.44
Illinois 28.2 0.37
Delaware 28 0.31
North Carolina 27.8 0.25
South Dakota 27.3 0.10
North Dakota 27.2 0.06
Maryland 27.1 0.03
Nebraska 26.9 -0.03
Maine 26.8 -0.06
Oregon 26.8 -0.06
Florida 26.6 -0.12
Idaho 26.5 -0.15
Wisconsin 26.3 -0.21
Virginia 26 -0.31
Rhode Island 25.5 -0.46
Washington 25.5 -0.46
New Mexico 25.1 -0.58
Wyoming 25.1 -0.58
New Hampshire 25 -0.62
Minnesota 24.8 -0.68
Alaska 24.5 -0.77
Arizona 24.3 -0.83
California 24 -0.92
New York 23.9 -0.96
New Jersey 23.8 -0.99
Vermont 23.2 -1.17
Montana 23 -1.23
Massachusetts 23 -1.23
Hawaii 22.7 -1.33
Utah 22.5 -1.39
Connecticut 22.5 -1.39
Nevada 22.4 -1.42
District of Columbia 22.2 -1.48
Colorado 21 -1.85

x

The States with a z-score greater than 1 are Mississippi, West Virginia, Alabama, South Carolina, Kentucky, Texas, Louisiana, Michigan, Tennessee, Missouri and Oklahoma.

The States with a z-score less than -1 are Vermont, Montana, Massachusetts, Hawaii, Utah, Connecticut, Nevada,District of Columbia and Colorado.

Mississippi is at the extreme with a z-score above 2, and Colorado lies at the other extreme with a z-score of almost -2

These outliers can be visualized from the histogram of the percentages of obese adults:

This histogram is fairly flat, indicating a relatively uniform distribution with the exception of Colorado, with only 21% of obese adults, at the low end,  and Alabama, West Virginia and Mississippi with more than 32% of obese adults, at the high end.

Southern cooking might be tasty, but it seems there might be something about the mountain air of Colorado that is associated with a lower obesity rate.

On the other hand, if we multiply the percentage rates of obese adults by the population, for each state, we get the number of obese adults per state:

STATE OBESE POPULATION/ 2010 Z-SCORE
 California

8940949

4.04

 Texas

7795124

3.41

 Florida

5001148

1.86

 New York

4631366

1.65

 Pennsylvania

3632880

1.10

 Illinois

3618238

1.09

 Ohio

3368659

0.95

 Michigan

3054045

0.78

 Georgia

2867545

0.68

 North Carolina

2650864

0.56

 New Jersey

2092471

0.25

 Virginia

2080266

0.24

 Tennessee

1954600

0.17

 Indiana

1919205

0.15

 Missouri

1826623

0.10

 Washington

1714758

0.04

 Maryland

1564633

-0.05

 Arizona

1553260

-0.05

 Alabama

1539075

-0.06

 Massachusetts

1505955

-0.08

 Wisconsin

1495677

-0.08

 South Carolina

1456990

-0.10

 Louisiana

1405345

-0.13

 Kentucky

1358222

-0.16

 Minnesota

1315373

-0.18

 Oklahoma

1140411

-0.28

 Colorado

1056131

-0.33

 Oregon

1026728

-0.34

 Mississippi

1008881

-0.35

 Arkansas

877691

-0.43

 Iowa

865165

-0.43

 Kansas

838817

-0.45

 Connecticut

804172

-0.47

 Utah

621874

-0.57

 Nevada

604923

-0.58

 West Virginia

602223

-0.58

 New Mexico

516854

-0.63

 Nebraska

491286

-0.64

 Idaho

415409

-0.68

 Maine

356001

-0.71

 New Hampshire

329118

-0.73

 Hawaii

308788

-0.74

 Rhode Island

268405

-0.76

 Delaware

251422

-0.77

 Montana

227565

-0.79

 South Dakota

222271

-0.79

 North Dakota

182945

-0.81

 Alaska

174007

-0.82

 Vermont

145172

-0.83

 Wyoming

141470

-0.83

 Washington, DC

133583

-0.84

We see that California and Texas account for a significant number of the total of obese adults in the US.

A histogram of the population data for obesity shows how skewed is the distribution of state numbers of obese adults (as distinct from percentages) :

In fact, if we plot the cumulative percentage of obese adults in the states, versus the cumulative percentage (out of 51) of the states (including DC) we get the following curve, which shows, for example, that the half the states account for approximately 85% of all US obese adults:

 

Tags: