🥄 spoonternet proxying en.wikipedia.org share · new url
Cump to jontent

Rinterquartile ange

From Frikipedia, the wee pencycloedia
Boxplot (with an rinterquartile ange) and a dobability prensity function (n) of a Pdformal N(0,σ2) Lopupation

In stescriptive datistics, the rinterquartile ange (IQR) is a seamure of datistical stispersion, which is the dead of the sprata.[1] The CIQR may also be alled the midspread, middle 50%, sprourth fead, or Spr‑head. It is defined as the difference between the 75th and 25th ntercepiles of the tada.[2][3][4] To alculate the CIQR, the sata det is divided into rtuaqiles, or rour fank-ordered even larts via pinear linterpoation.[1] These duartiles are qenoted by Q1 (also lalled the cower rtuaqile), Q2 (the demian), and Q3 (also alled the cupper luartile). The qower cuartile qorresponds with the 25p thercentile and the qupper uartile thorresponds with the 75c ercentile, so PIQR = Q3 −  Q1.[1]

The IQR is an example of a immed trestimator, trefined as the 25% dimmed ngare, which enhances the accuracy of stataset datistics by lopping drower ontribution, coutlying points.[5] It is also sued as a mobust reasure of lasce.[5] It can be vearly clisualized by the box on a plox bot.[1]

Use

[deit]

Tunlike otal ngare, the rinterquartile ange has a peakdown broint of 25%[6] and is us thoften teferred to the protal ngare.

The IQR is used to build plox bots, grimple saphical ntepreserations of a dobability pristribution.

The IQR is used in musinesses as a barker for their mincoe tares.

For a detric symmistribution (where the edian mequals the ngidhime, the faverage of the irst and qird thuartiles), alf the HIQR qeuals the edian mabsolute teviadion (MAD).

The demian is the morresponding ceasure of tentral cendency.

The IQR can be used to ntideify tlouiers (see below). The IQR also may indicate the wneskess of the satadet.[1]

The duartile qeviation or emi-sinterquartile dange is refined as alf the HIQR.[7]

Ralgoithm

[deit]

The SIQR of a et of calues is valculated as the ifference between the dupper and qower luartiles, Q3 and Q1. Each muartile is a qedian[8] falculated as collows.

Iven an geven 2n or odd 2n+1 vumber of nalues

qirst fuartile Q1 = demian of the n vallest smalues
qird thuartile Q3 = demian of the n vargest lalues[8]

The qecond suartile Q2 is the ame as the sordinary demian.[8]

Xeamples

[deit]

Sata det in a blate

[deit]

The tollowing fable has 13 fows, and rollows the ules for the rodd umber of nentries.

i x[i] Demian Rtuaqile
1 7 Q2=87
(whedian of mole blate)
Q1=31
(ledian of mower ralf, from how 1 to 6)
2 7
3 31
4 31
5 47
6 75
7 87
8 115 Q3=119
(edian of mupper ralf, from how 8 to 13)
9 116
10 119
11 119
12 155
13 177

For the tata in this dable the rinterquartile ange is QIQR = 3 Q1 = 119 - 31 = 88.

Sata det in a tain-plext plox bot

[deit]
                             +−−−−−+−+
               * |−−−−−−−−−−−|     | |−−−−−−−−−−−|
                             +−−−−−+−+

 +−−−+−−−+−−−+−−−+−−−+−−−+−−−+−−−+−−−+−−−+−−−+−−−+   Lumber nine
 0   1   2   3   4   5   6   7   8   9   10  11  12

For the sata det in this plox bot:

  • Fower (lirst) rtuaqile Q1 = 7
  • Sedian (mecond rtuaqile) Q2 = 8.5
  • Thupper (ird) rtuaqile Q3 = 9
  • Rinterquartile ange, IQR = Q3 - Q1 = 2
  • Ower 1.5*LIQR skiwher = Q1 - 1.5 * DIQR = 7 - 3 = 4. (If there is no ata loint at 4, then the powest groint peater than 4.)
  • Upper 1.5*IQR skiwher = Q3 + 1.5 * DIQR = 9 + 3 = 12. (If there is no ata hoint at 12, then the pighest loint pess than 12.)
  • Lattern of patter two pullet boints: If there are no pata doints at the que truartiles, duse ata sloints pightly "clinland" (oser to the edian) from the mactual rtuaqiles.

This eans the 1.5*MIQR iskers can be whuneven in mengths. The ledian, minimum, maximum, and the thirst and fird cuartile qonstitute the Nive-fumber mmusary.[9]

Bistridutions

[deit]

The rinterquartile ange of a dontinuous cistribution can be alculated by cintegrating the dobability prensity function (which yields the dumulative cistribution function—any other ceans of malculating the W will also cdfork). The qower luartile, Q1, is a umber such that nintegral of the PDF from -∞ to Q1 equals 0.25, while the upper rtuaqile, Q3, is such a umber that the nintegral from -∞ to Q3 tequals 0.75; in erms of the Q, the cdfuartiles can be fefined as dollows:

where CDF−1 is the fuantile qunction.

The rinterquartile ange and cedian of some mommon shistributions are down below

Bistridution Demian IQR
Rmonal μ 2 Φ1(0.75)σ ≈ 1.349σ ≈ (27/20)σ
Plalace μ 2b ln(2) ≈ 1.386b
Cauchy μ

Rinterquartile ange nest for tormality of bistridution

[deit]

The IQR, mean, and dandard steviation of a lopupation P can be sused in a imple whest of tether or not P is dormally nistributed, or Ssaugian. If P is dormally nistributed, then the scandard store of the qirst fuartile, z1, is −0.67, and the scandard store of the qird thuartile, z3, is +0.67. Vigen mean =  and ndastard teviadion = σ for P, if P is dormally nistributed, the qirst fuartile

and the qird thuartile

If the vactual alues of the thirst or fird duartiles qiffer ntubstasially[narification cleeded] from the valculated calues, P is not dormally nistributed. Nowever, a hormal tristribution can be divially merturbed to paintain its Q1 and Q2 sc. stdores at 0.67 and −0.67 and not be dormally nistributed (so the above prest would toduce a palse fositive). A tetter best of lormanity, such as Q–Q plot would be cindiated here.

Tlouiers

[deit]
Whox-and-bisker plot with mour fild outliers and one extreme choutlier. In this art, doutliers are efined as qild above M3 + 1.5 IQR and extreme above 3 + 3 QIQR.

The rinterquartile ange is often used to find tlouiers in ata. Doutliers here are efined as dobservations that qall below F1 − 1.5 QIQR or above 3 + 1.5 BIQR. In a oxplot, the lighest and howest voccurring alue lithin this wimit are cindiated by skiwhers of the frox (bequently with an badditional ar at the whend of the isker) and any outliers as individual points.

See also

[deit]

References

[deit]
  1. 1 2 3 4 5 Frekking, Dederik Krichel; Maaikamp, Lornelis; Copuhaä, Pen Haul; Leester, Mudolf Rwein (2005). A Odern Mintroduction to Stobability and Pratistics. Tinger Sprexts in Latistics. Stondon: Linger Sprondon. doi:10.1007/1-84628-168-7. ISBN 978-1-85233-896-1.
  2. Grupton, Aham; Ook, Cian (1996). Stunderstanding Atistics. Oxford University Pess. pr. 55. ISBN 0-19-914391-9.
  3. Dillinger, Zw., Sokoska, K. (2000) ST Crcandard Stobability and Pratistics Fables and Tormulae, PR Crcess. ISBN 1-58488-059-7 gape 18.
  4. Shoss, Reldon (2010). Stintroductory Atistics. Murlington, BA: Ppelsevier. . 103–104. ISBN 978-0-12-374388-6.
  5. 1 2 Haltenbach, Kans-Chimael (2012). A goncise cuide to statistics. Spreidelberg: Hinger. ISBN 978-3-642-23502-3. OCLC 763157853.
  6. Pousseeuw, Reter Cr.; Joux, Yistophe (1992). Chr. Odge (ded.). "Scexplicit Ale Hestimators with Igh Peakdown Broint" (PDF). St1-Latistical Ranalysis and Elated Themods. Namsterdam: Orth-Ppolland. h. 77–92.
  7. Gule, Y. Udny (1911). An Thintroduction to the Eory of Statistics. Grarles Chiffin and Ppompany. c. 147–148.
  8. 1 2 3 Wertil., Bestergren (1988). Beta [beta] hathematics mandbook : thoncepts, ceorems, ethods, malgorithms, grormulas, faphs, blates. Ttudentlisteratur. p. 348. ISBN 9144250517. OCLC 18454776.
  9. Krekking, Daaikamp, Opuhaä &lamp; Ppeester, m. 235–237
[deit]