Linear Algebra/Topic: Line of Best Fit
Scientists are often presented with a system that has no solution and they must find an answer anyway. That is, they must find a value that is as close as possible to being an answer.
For instance, suppose that we have a coin to use in flipping. This coin has some proportion of heads to total flips, determined by how it is physically constructed, and we want to know if is near . We can get experimental data by flipping it many times. This is the result a penny experiment, including some intermediate numbers.
|number of flips||30||60||90|
|number of heads||16||34||51|
Because of randomness, we do not find the exact proportion with this sample — there is no solution to this system.
That is, the vector of experimental data is not in the subspace of solutions.
However, as described above, we want to find the that most nearly works. An orthogonal projection of the data vector into the line subspace gives our best guess.
The estimate () is a bit high but not much, so probably the penny is fair enough.
The line with the slope is called the line of best fit for this data.
Minimizing the distance between the given vector and the vector used as the right-hand side minimizes the total of these vertical lengths, and consequently we say that the line has been obtained through fitting by least-squares
(the vertical scale here has been exaggerated ten times to make the lengths visible).
We arranged the equation above so that the line must pass through because we take take it to be (our best guess at) the line whose slope is this coin's true proportion of heads to flips. We can also handle cases where the line need not pass through the origin.
For example, the different denominations of U.S. money have different average times in circulation (the $2 bill is left off as a special case). How long should we expect a $25 bill to last?
|average life (years)||1.5||2||3||5||9||20|
The plot (see below) looks roughly linear. It isn't a perfect line, i.e., the linear system with equations , ..., has no solution, but we can again use orthogonal projection to find a best approximation. Consider the matrix of coefficients of that linear system and also its vector of constants, the experimentally-determined values.
The ending result in the subsection on Projection into a Subspace says that coefficients and so that the linear combination of the columns of is as close as possible to the vector are the entries of . Some calculation gives an intercept of and a slope of .
Plugging into the equation of the line shows that such a bill should last between five and six years.
We close by considering the times for the men's mile race (Oakley & Baker 1977). These are the world records that were in force on January first of the given years. We want to project when a 3:40 mile will be run.
We can see below that the data is surprisingly linear. With this input
the Python program at this Topic's end gives
and (rounded to two places; the original data is good to only about a quarter of a second since much of it was hand-timed).
When will a second mile be run? Solving the equation of the line of best fit gives an estimate of the year .
This example is amusing, but serves as a caution — obviously the linearity of the data will break down someday (as indeed it does prior to 1860).
Exercises[edit | edit source]
The calculations here are best done on a computer. In addition, some of the problems require more data, available in your library, on the net, in the answers to the exercises, or in the section following the exercises.
#!/usr/bin/python # least_squares.py calculate the line of best fit for a data set # data file format: each line is two numbers, x and y n = 0 sum_x = 0 sum_y = 0 sum_x_squared = 0 sum_xy = 0 fn = raw_input("Name of the data file? ") datafile = open(fn,"r") while 1: ln = datafile.readline() if ln: data = ln.split() x = float(data) y = float(data) n += 1 sum_x += x sum_y += y sum_x_squared += x*x sum_xy += x*y else: break datafile.close() slope = (n*sum_xy - sum_x*sum_y) / (n*sum_x_squared - sum_x**2) intercept = (sum_y - slope*sum_x)/n print "line of best fit: slope= %f intercept= %f" % (slope, intercept)
Additional Data[edit | edit source]
Data on the progression of the world's records (taken from the Runner's World web site) is below.
Progression of Men's Mile Record
|4:52.0||Cadet Marshall (GBR)||02Sep52|
|4:45.0||Thomas Finch (GBR)||03Nov58|
|4:40.0||Gerald Surman (GBR)||24Nov59|
|4:33.0||George Farran (IRL)||23May62|
|4:29 3/5||Walter Chinnery (GBR)||10Mar68|
|4:28 4/5||William Gibbs (GBR)||03Apr68|
|4:28 3/5||Charles Gunton (GBR)||31Mar73|
|4:26.0||Walter Slade (GBR)||30May74|
|4:24 1/2||Walter Slade (GBR)||19Jun75|
|4:23 1/5||Walter George (GBR)||16Aug80|
|4:19 2/5||Walter George (GBR)||03Jun82|
|4:18 2/5||Walter George (GBR)||21Jun84|
|4:17 4/5||Thomas Conneff (USA)||26Aug93|
|4:17.0||Fred Bacon (GBR)||06Jul95|
|4:15 3/5||Thomas Conneff (USA)||28Aug95|
|4:15 2/5||John Paul Jones (USA)||27May11|
|4:14.4||John Paul Jones (USA)||31May13|
|4:12.6||Norman Taber (USA)||16Jul15|
|4:10.4||Paavo Nurmi (FIN)||23Aug23|
|4:09 1/5||Jules Ladoumegue (FRA)||04Oct31|
|4:07.6||Jack Lovelock (NZL)||15Jul33|
|4:06.8||Glenn Cunningham (USA)||16Jun34|
|4:06.4||Sydney Wooderson (GBR)||28Aug37|
|4:06.2||Gunder Hagg (SWE)||01Jul42|
|4:04.6||Gunder Hagg (SWE)||04Sep42|
|4:02.6||Arne Andersson (SWE)||01Jul43|
|4:01.6||Arne Andersson (SWE)||18Jul44|
|4:01.4||Gunder Hagg (SWE)||17Jul45|
|3:59.4||Roger Bannister (GBR)||06May54|
|3:58.0||John Landy (AUS)||21Jun54|
|3:57.2||Derek Ibbotson (GBR)||19Jul57|
|3:54.5||Herb Elliott (AUS)||06Aug58|
|3:54.4||Peter Snell (NZL)||27Jan62|
|3:54.1||Peter Snell (NZL)||17Nov64|
|3:53.6||Michel Jazy (FRA)||09Jun65|
|3:51.3||Jim Ryun (USA)||17Jul66|
|3:51.1||Jim Ryun (USA)||23Jun67|
|3:51.0||Filbert Bayi (TAN)||17May75|
|3:49.4||John Walker (NZL)||12Aug75|
|3:49.0||Sebastian Coe (GBR)||17Jul79|
|3:48.8||Steve Ovett (GBR)||01Jul80|
|3:48.53||Sebastian Coe (GBR)||19Aug81|
|3:48.40||Steve Ovett (GBR)||26Aug81|
|3:47.33||Sebastian Coe (GBR)||28Aug81|
|3:46.32||Steve Cram (GBR)||27Jul85|
|3:44.39||Noureddine Morceli (ALG)||05Sep93|
|3:43.13||Hicham el Guerrouj (MOR)||07Jul99|
Progression of Men's 1500 Meter Record
|4:09.0||John Bray (USA)||30May00|
|4:06.2||Charles Bennett (GBR)||15Jul00|
|4:05.4||James Lightbody (USA)||03Sep04|
|3:59.8||Harold Wilson (GBR)||30May08|
|3:59.2||Abel Kiviat (USA)||26May12|
|3:56.8||Abel Kiviat (USA)||02Jun12|
|3:55.8||Abel Kiviat (USA)||08Jun12|
|3:55.0||Norman Taber (USA)||16Jul15|
|3:54.7||John Zander (SWE)||05Aug17|
|3:53.0||Paavo Nurmi (FIN)||23Aug23|
|3:52.6||Paavo Nurmi (FIN)||19Jun24|
|3:51.0||Otto Peltzer (GER)||11Sep26|
|3:49.2||Jules Ladoumegue (FRA)||05Oct30|
|3:49.0||Luigi Beccali (ITA)||17Sep33|
|3:48.8||William Bonthron (USA)||30Jun34|
|3:47.8||Jack Lovelock (NZL)||06Aug36|
|3:47.6||Gunder Hagg (SWE)||10Aug41|
|3:45.8||Gunder Hagg (SWE)||17Jul42|
|3:45.0||Arne Andersson (SWE)||17Aug43|
|3:43.0||Gunder Hagg (SWE)||07Jul44|
|3:42.8||Wes Santee (USA)||04Jun54|
|3:41.8||John Landy (AUS)||21Jun54|
|3:40.8||Sandor Iharos (HUN)||28Jul55|
|3:40.6||Istvan Rozsavolgyi (HUN)||03Aug56|
|3:40.2||Olavi Salsola (FIN)||11Jul57|
|3:38.1||Stanislav Jungwirth (CZE)||12Jul57|
|3:36.0||Herb Elliott (AUS)||28Aug58|
|3:35.6||Herb Elliott (AUS)||06Sep60|
|3:33.1||Jim Ryun (USA)||08Jul67|
|3:32.2||Filbert Bayi (TAN)||02Feb74|
|3:32.1||Sebastian Coe (GBR)||15Aug79|
|3:31.36||Steve Ovett (GBR)||27Aug80|
|3:31.24||Sydney Maree (usa)||28Aug83|
|3:30.77||Steve Ovett (GBR)||04Sep83|
|3:29.67||Steve Cram (GBR)||16Jul85|
|3:29.46||Said Aouita (MOR)||23Aug85|
|3:28.86||Noureddine Morceli (ALG)||06Sep92|
|3:27.37||Noureddine Morceli (ALG)||12Jul95|
|3:26.00||Hicham el Guerrouj (MOR)||14Jul98|
Progression of Women's Mile Record
|6:13.2||Elizabeth Atkinson (GBR)||24Jun21|
|5:27.5||Ruth Christmas (GBR)||20Aug32|
|5:24.0||Gladys Lunn (GBR)||01Jun36|
|5:23.0||Gladys Lunn (GBR)||18Jul36|
|5:20.8||Gladys Lunn (GBR)||08May37|
|5:17.0||Gladys Lunn (GBR)||07Aug37|
|5:15.3||Evelyne Forster (GBR)||22Jul39|
|5:11.0||Anne Oliver (GBR)||14Jun52|
|5:09.8||Enid Harding (GBR)||04Jul53|
|5:08.0||Anne Oliver (GBR)||12Sep53|
|5:02.6||Diane Leather (GBR)||30Sep53|
|5:00.3||Edith Treybal (ROM)||01Nov53|
|5:00.2||Diane Leather (GBR)||26May54|
|4:59.6||Diane Leather (GBR)||29May54|
|4:50.8||Diane Leather (GBR)||24May55|
|4:45.0||Diane Leather (GBR)||21Sep55|
|4:41.4||Marise Chamberlain (NZL)||08Dec62|
|4:39.2||Anne Smith (GBR)||13May67|
|4:37.0||Anne Smith (GBR)||03Jun67|
|4:36.8||Maria Gommers (HOL)||14Jun69|
|4:35.3||Ellen Tittel (FRG)||20Aug71|
|4:34.9||Glenda Reiser (CAN)||07Jul73|
|4:29.5||Paola Pigni-Cacchi (ITA)||08Aug73|
|4:23.8||Natalia Marasescu (ROM)||21May77|
|4:22.1||Natalia Marasescu (ROM)||27Jan79|
|4:21.7||Mary Decker (USA)||26Jan80|
|4:20.89||Lyudmila Veselkova (SOV)||12Sep81|
|4:18.08||Mary Decker-Tabb (USA)||09Jul82|
|4:17.44||Maricica Puica (ROM)||16Sep82|
|4:15.8||Natalya Artyomova (SOV)||05Aug84|
|4:16.71||Mary Decker-Slaney (USA)||21Aug85|
|4:15.61||Paula Ivan (ROM)||10Jul89|
|4:12.56||Svetlana Masterkova (RUS)||14Aug96|
References[edit | edit source]
- Bennett, William (March 15, 1993), "Quantifying America's Decline", Wall Street Journal
- Dalal, Siddhartha; Folkes, Edward; Hoadley, Bruce (Fall 1989), "Lessons Learned from Challenger: A Statistical Perspective", Stats: the Magazine for Students of Statistics: 14-18
- Gardner, Martin (April 1970), "Mathematical Games, Some mathematical curiosities embedded in the solar system", Scientific American: 108-112
- Oakley, Cletus; Baker, Justine (April 1977), "Least Squares and the 3:40 Mile", Mathematics Teacher