Movatterモバイル変換

[0]ホーム
URL:

Jump to content
WikipediaThe Free Encyclopedia
Search

Student'st-test

From Wikipedia, the free encyclopedia
(Redirected fromT test)
Statistical hypothesis test

Student'st-test is astatistical test used to test whether the difference between the response of two groups isstatistically significant or not. It is anystatistical hypothesis test in which thetest statistic follows aStudent'st-distribution under thenull hypothesis. It is most commonly applied when the test statistic would follow anormal distribution if the value of ascaling term in the test statistic were known (typically, the scaling term is unknown and is therefore anuisance parameter). When the scaling term is estimated based on thedata, the test statistic—under certain conditions—follows a Student'st distribution. Thet-test's most common application is to test whether the means of two populations are significantly different. In many cases, aZ-test will yield very similar results to at-test because the latter converges to the former as the size of the dataset increases.

History

[edit]
William Sealy Gosset, who developed the "t-statistic" and published it under thepseudonym of "Student"

The term "t-statistic" is abbreviated from "hypothesis test statistic".[1] In statistics, thet-distribution was first derived as aposterior distribution in 1876 byHelmert[2][3][4] andLüroth.[5][6][7] Thet-distribution also appeared in a more general form as Pearson type IV distribution inKarl Pearson's 1895 paper.[8] However, thet-distribution, also known asStudent'st-distribution, gets its name fromWilliam Sealy Gosset, who first published it in English in 1908 in the scientific journalBiometrika using the pseudonym "Student"[9][10] because his employer preferred staff to usepen names when publishing scientific papers.[11] Gosset worked at theGuinness Brewery inDublin,Ireland, and was interested in the problems of small samples – for example, the chemical properties of barley with small sample sizes. Hence a second version of the etymology of the term Student is that Guinness did not want their competitors to know that they were using thet-test to determine the quality of raw material. Although it was William Gosset after whom the term "Student" is penned, it was actually through the work ofRonald Fisher that the distribution became well known as "Student's distribution"[12] and "Student'st-test".

Gosset devised thet-test as an economical way to monitor the quality ofstout. Thet-test work was submitted to and accepted in the journalBiometrika and published in 1908.[9]

Guinness had a policy of allowing technical staff leave for study (so-called "study leave"), which Gosset used during the first two terms of the 1906–1907 academic year in ProfessorKarl Pearson's Biometric Laboratory atUniversity College London.[13] Gosset's identity was then known to fellow statisticians and to editor-in-chief Karl Pearson.[14]

Uses

[edit]

One-samplet-test

[edit]

Aone-sample Student'st-test is alocation test of whether the mean of a population has a value specified in anull hypothesis. In testing the null hypothesis that the population mean is equal to a specified valueμ0, one uses the statistic

t=x¯−μ0s/n,{\displaystyle t={\frac {{\bar {x}}-\mu _{0}}{s/{\sqrt {n}}}},}

wherex¯{\displaystyle {\bar {x}}} is the sample mean,s is thesample standard deviation andn is the sample size. Thedegrees of freedom used in this test aren − 1. Although the parent population does not need to be normally distributed, the distribution of the population of sample meansx¯{\displaystyle {\bar {x}}} is assumed to be normal.

By thecentral limit theorem, if the observations are independent and the second moment exists, thent{\displaystyle t} will be approximately normalN(0,1){\textstyle {\mathcal {N}}(0,1)}.

Two-samplet-tests

[edit]
Type I error of unpaired and paired two-samplet-tests as a function of the correlation. The simulated random numbers originate from a bivariate normal distribution with a variance of 1. The significance level is 5% and the number of cases is 60.
Power of unpaired and paired two-samplet-tests as a function of the correlation. The simulated random numbers originate from a bivariate normal distribution with a variance of 1 and a deviation of the expected value of 0.4. The significance level is 5% and the number of cases is 60.

Atwo-sample location test of the null hypothesis such that themeans of two populations are equal. All such tests are usually calledStudent'st-tests, though strictly speaking that name should only be used if thevariances of the two populations are also assumed to be equal; the form of the test used when this assumption is dropped is sometimes calledWelch'st-test. These tests are often referred to asunpaired orindependent samplest-tests, as they are typically applied when thestatistical units underlying the two samples being compared are non-overlapping.[15]

Two-samplet-tests for a difference in means involve independent samples (unpaired samples) orpaired samples. Pairedt-tests are a form ofblocking, and have greaterpower (probability of avoiding a type II error, also known as a false negative) than unpaired tests when the paired units are similar with respect to "noise factors" (seeconfounder) that are independent of membership in the two groups being compared.[16] In a different context, pairedt-tests can be used to reduce the effects ofconfounding factors in anobservational study.

Independent (unpaired) samples

[edit]

The independent samplest-test is used when two separate sets ofindependent and identically distributed samples are obtained, and one variable from each of the two populations is compared. For example, suppose we are evaluating the effect of a medical treatment, and we enroll 100 subjects into our study, then randomly assign 50 subjects to the treatment group and 50 subjects to the control group. In this case, we have two independent samples and would use the unpaired form of thet-test.

Paired samples

[edit]
Main article:Paired difference test

Paired samplest-tests typically consist of a sample of matched pairs of similarunits, or one group of units that has been tested twice (a "repeated measures"t-test).

A typical example of the repeated measurest-test would be where subjects are tested prior to a treatment, say for high blood pressure, and the same subjects are tested again after treatment with a blood-pressure-lowering medication. By comparing the same patient's numbers before and after treatment, we are effectively using each patient as their own control. That way the correct rejection of the null hypothesis (here: of no difference made by the treatment) can become much more likely, with statistical power increasing simply because the random interpatient variation has now been eliminated. However, an increase of statistical power comes at a price: more tests are required, each subject having to be tested twice. Because half of the sample now depends on the other half, the paired version of Student'st-test has only⁠n/2⁠ − 1 degrees of freedom (withn being the total number of observations). Pairs become individual test units, and the sample has to be doubled to achieve the same number of degrees of freedom. Normally, there aren − 1 degrees of freedom (withn being the total number of observations).[17]

A paired samplest-test based on a "matched-pairs sample" results from an unpaired sample that is subsequently used to form a paired sample, by using additional variables that were measured along with the variable of interest.[18] The matching is carried out by identifying pairs of values consisting of one observation from each of the two samples, where the pair is similar in terms of other measured variables. This approach is sometimes used in observational studies to reduce or eliminate the effects of confounding factors.

Paired samplest-tests are often referred to as "dependent samplest-tests".

Assumptions

[edit]

[dubious –discuss]

Most test statistics have the formt =Z/s, whereZ ands are functions of the data.

Z may be sensitive to the alternative hypothesis (i.e., its magnitude tends to be larger when the alternative hypothesis is true), whereass is ascaling parameter that allows the distribution oft to be determined.

As an example, in the one-samplet-test

t=Zs=X¯−μσ^/n,{\displaystyle t={\frac {Z}{s}}={\frac {{\bar {X}}-\mu }{{\hat {\sigma }}/{\sqrt {n}}}},}

whereX¯{\displaystyle {\bar {X}}} is thesample mean from a sampleX1,X2, …,Xn, of sizen,s is thestandard error of the mean,σ^=1n−1∑i(Xi−X¯)2{\displaystyle {\hat {\sigma }}={\sqrt {{\frac {1}{n-1}}\sum _{i}(X_{i}-{\bar {X}})^{2}}}} is the estimate of thestandard deviation of the population, andμ is thepopulation mean.

The assumptions underlying at-test in the simplest form above are that:

  • X follows a normal distribution with meanμ and varianceσ2/n.
  • s2(n − 1)/σ2 follows aχ2 distribution withn − 1degrees of freedom. This assumption is met when the observations used for estimatings2 come from a normal distribution (andi.i.d. for each group).
  • Z ands areindependent.

In thet-test comparing the means of two independent samples, the following assumptions should be met:

  • The means of the two populations being compared should follownormal distributions. Under weak assumptions, this follows in large samples from thecentral limit theorem, even when the distribution of observations in each group is non-normal.[19]
  • If using Student's original definition of thet-test, the two populations being compared should have the same variance (testable usingF-test,Levene's test,Bartlett's test, or theBrown–Forsythe test; or assessable graphically using aQ–Q plot). If the sample sizes in the two groups being compared are equal, Student's originalt-test is highly robust to the presence of unequal variances.[20]Welch'st-test is insensitive to equality of the variances regardless of whether the sample sizes are similar.
  • The data used to carry out the test should either be sampled independently from the two populations being compared or be fully paired. This is in general not testable from the data, but if the data are known to be dependent (e.g. paired by test design), a dependent test has to be applied. For partially paired data, the classical independentt-tests may give invalid results as the test statistic might not follow at distribution, while the dependentt-test is sub-optimal as it discards the unpaired data.[21]

Most two-samplet-tests are robust to all but large deviations from the assumptions.[22]

Forexactness, thet-test andZ-test require normality of the sample means, and thet-test additionally requires that the sample variance follows a scaledχ2 distribution, and that the sample mean and sample variance bestatistically independent. Normality of the individual data values is not required if these conditions are met. By thecentral limit theorem, sample means of moderately large samples are often well-approximated by a normal distribution even if the data are not normally distributed. However, the sample size required for the sample means to converge to normality depends on the skewness of the distribution of the original data. The sample can vary from 30 to 100 or higher values depending on the skewness.[23][24]

For non-normal data, the distribution of the sample variance may deviate substantially from aχ2 distribution.

However, if the sample size is large,Slutsky's theorem implies that the distribution of the sample variance has little effect on the distribution of the test statistic. That is, as sample sizen{\displaystyle n} increases:

n(X¯−μ)→dN(0,σ2){\displaystyle {\sqrt {n}}({\bar {X}}-\mu )\xrightarrow {d} N(0,\sigma ^{2})} as per theCentral limit theorem,
s2→pσ2{\displaystyle s^{2}\xrightarrow {p} \sigma ^{2}} as per thelaw of large numbers,
∴n(X¯−μ)s→dN(0,1){\displaystyle \therefore {\frac {{\sqrt {n}}({\bar {X}}-\mu )}{s}}\xrightarrow {d} N(0,1)}.

Calculations

[edit]

Explicit expressions that can be used to carry out varioust-tests are given below. In each case, the formula for a test statistic that either exactly follows or closely approximates at-distribution under the null hypothesis is given. Also, the appropriatedegrees of freedom are given in each case. Each of these statistics can be used to carry out either aone-tailed or two-tailed test.

Once thet value and degrees of freedom are determined, ap-value can be found using atable of values from Student'st-distribution. If the calculatedp-value is below the threshold chosen forstatistical significance (usually the 0.10, the 0.05, or 0.01 level), then the null hypothesis is rejected in favor of the alternative hypothesis.

Slope of a regression line

[edit]

Suppose one is fitting the model

Y=α+βx+ε,{\displaystyle Y=\alpha +\beta x+\varepsilon ,}

wherex is known,α andβ are unknown,ε is a normally distributed random variable with mean 0 and unknown varianceσ2, andY is the outcome of interest. We want to test the null hypothesis that the slopeβ is equal to some specified valueβ0 (often taken to be 0, in which case the null hypothesis is thatx andy are uncorrelated).

Let

α^,β^=least-squares estimators,SEα^,SEβ^=the standard errors of least-squares estimators.{\displaystyle {\begin{aligned}{\hat {\alpha }},{\hat {\beta }}&={\text{least-squares estimators}},\\SE_{\hat {\alpha }},SE_{\hat {\beta }}&={\text{the standard errors of least-squares estimators}}.\end{aligned}}}

Then

tscore=β^−β0SEβ^∼Tn−2{\displaystyle t_{\text{score}}={\frac {{\hat {\beta }}-\beta _{0}}{SE_{\hat {\beta }}}}\sim {\mathcal {T}}_{n-2}}

has at-distribution withn − 2 degrees of freedom if the null hypothesis is true. Thestandard error of the slope coefficient:

SEβ^=1n−2∑i=1n(yi−y^i)2∑i=1n(xi−x¯)2{\displaystyle SE_{\hat {\beta }}={\frac {\sqrt {\displaystyle {\frac {1}{n-2}}\sum _{i=1}^{n}(y_{i}-{\hat {y}}_{i})^{2}}}{\sqrt {\displaystyle \sum _{i=1}^{n}(x_{i}-{\bar {x}})^{2}}}}}

can be written in terms of the residuals. Let

ε^i=yi−y^i=yi−(α^+β^xi)=residuals=estimated errors,SSR=∑i=1nε^i2=sum of squares of residuals.{\displaystyle {\begin{aligned}{\hat {\varepsilon }}_{i}&=y_{i}-{\hat {y}}_{i}=y_{i}-({\hat {\alpha }}+{\hat {\beta }}x_{i})={\text{residuals}}={\text{estimated errors}},\\{\text{SSR}}&=\sum _{i=1}^{n}{{\hat {\varepsilon }}_{i}}^{2}={\text{sum of squares of residuals}}.\end{aligned}}}

Thentscore is given by

tscore=(β^−β0)n−2SSR∑i=1n(xi−x¯)2.{\displaystyle t_{\text{score}}={\frac {({\hat {\beta }}-\beta _{0}){\sqrt {n-2}}}{\sqrt {\frac {SSR}{\sum _{i=1}^{n}(x_{i}-{\bar {x}})^{2}}}}}.}

Another way to determine thetscore is

tscore=rn−21−r2,{\displaystyle t_{\text{score}}={\frac {r{\sqrt {n-2}}}{\sqrt {1-r^{2}}}},}

wherer is thePearson correlation coefficient.

Thetscore, intercept can be determined from thetscore, slope:

tscore,intercept=αβtscore,slopesx2+x¯2,{\displaystyle t_{\text{score,intercept}}={\frac {\alpha }{\beta }}{\frac {t_{\text{score,slope}}}{\sqrt {s_{\text{x}}^{2}+{\bar {x}}^{2}}}},}

wheresx2 is the sample variance.

Independent two-samplet-test

[edit]

Equal sample sizes and variance

[edit]

Given two groups (1, 2), this test is only applicable when:

  • the two sample sizes are equal,
  • it can be assumed that the two distributions have the same variance.

Violations of these assumptions are discussed below.

Thet statistic to test whether the means are different can be calculated as follows:

t=X¯1−X¯2sp2n,{\displaystyle t={\frac {{\bar {X}}_{1}-{\bar {X}}_{2}}{s_{p}{\sqrt {\frac {2}{n}}}}},}

where

sp=sX12+sX222.{\displaystyle s_{p}={\sqrt {\frac {s_{X_{1}}^{2}+s_{X_{2}}^{2}}{2}}}.}

Heresp is thepooled standard deviation forn =n1 =n2, ands 2
X1 ands 2
X2 are theunbiased estimators of the population variance. The denominator oft is thestandard error of the difference between two means.

For significance testing, thedegrees of freedom for this test is2n − 2, wheren is sample size.

Equal or unequal sample sizes, similar variances (⁠1/2⁠ <⁠sX1/sX2⁠ < 2)

[edit]

This test is used only when it can be assumed that the two distributions have the same variance (when this assumption is violated, see below). The previous formulae are a special case of the formulae below, one recovers them when both samples are equal in size:n =n1 =n2.

Thet statistic to test whether the means are different can be calculated as follows:

t=X¯1−X¯2sp⋅1n1+1n2,{\displaystyle t={\frac {{\bar {X}}_{1}-{\bar {X}}_{2}}{s_{p}\cdot {\sqrt {{\frac {1}{n_{1}}}+{\frac {1}{n_{2}}}}}}},}

where

sp=(n1−1)sX12+(n2−1)sX22n1+n2−2{\displaystyle s_{p}={\sqrt {\frac {(n_{1}-1)s_{X_{1}}^{2}+(n_{2}-1)s_{X_{2}}^{2}}{n_{1}+n_{2}-2}}}}

is thepooled standard deviation of the two samples: it is defined in this way so that its square is anunbiased estimator of the common variance, whether or not the population means are the same. In these formulae,ni − 1 is the number of degrees of freedom for each group, and the total sample size minus two (that is,n1 + n2 − 2) is the total number of degrees of freedom, which is used in significance testing.

Theminimum detectable effect (MDE) is:[25]

δ≥2Sp2n(t1−α,ν+t1−β,ν){\displaystyle \delta \geq {\sqrt {\frac {2S_{p}^{2}}{n}}}(t_{1-\alpha ,\nu }+t_{1-\beta ,\nu })}

Equal or unequal sample sizes, unequal variances (sX1 > 2sX2 orsX2 > 2sX1)

[edit]
Main article:Welch'st-test

This test, also known as Welch'st-test, is used only when the two population variances are not assumed to be equal (the two sample sizes may or may not be equal) and hence must be estimated separately. Thet statistic to test whether the population means are different is calculated as

t=X¯1−X¯2sΔ¯,{\displaystyle t={\frac {{\bar {X}}_{1}-{\bar {X}}_{2}}{s_{\bar {\Delta }}}},}

where

sΔ¯=s12n1+s22n2.{\displaystyle s_{\bar {\Delta }}={\sqrt {{\frac {s_{1}^{2}}{n_{1}}}+{\frac {s_{2}^{2}}{n_{2}}}}}.}

Heresi2 is theunbiased estimator of thevariance of each of the two samples withni = number of participants in groupi (i = 1 or 2). In this case(sΔ¯)2{\displaystyle (s_{\bar {\Delta }})^{2}}is not a pooled variance. For use in significance testing, the distribution of the test statistic is approximated as an ordinary Student'st-distribution with the degrees of freedom calculated using

d.f.=(s12n1+s22n2)2(s12/n1)2n1−1+(s22/n2)2n2−1.{\displaystyle {\text{d.f.}}={\frac {\left({\frac {s_{1}^{2}}{n_{1}}}+{\frac {s_{2}^{2}}{n_{2}}}\right)^{2}}{{\frac {(s_{1}^{2}/n_{1})^{2}}{n_{1}-1}}+{\frac {(s_{2}^{2}/n_{2})^{2}}{n_{2}-1}}}}.}

This is known as theWelch–Satterthwaite equation. The true distribution of the test statistic actually depends (slightly) on the two unknown population variances (seeBehrens–Fisher problem).

Exact method for unequal variances and sample sizes

[edit]

The test[26] deals with the famousBehrens–Fisher problem, i.e., comparing the difference between the means of two normally distributed populations when the variances of the two populations are not assumed to be equal, based on two independent samples.

The test is developed as anexact test that allows forunequal sample sizes andunequal variances of two populations. The exact property still holds even withextremely small andunbalanced sample sizes (e.g. m≡nX=50 {\displaystyle \ m\equiv n_{\mathsf {X}}=50\ } vs. n≡nY=5 {\displaystyle \ n\equiv n_{\mathsf {Y}}=5\ }).

The statistic to test whether the means are different can be calculated as follows:

Let X=[ X1,X2,…,Xm ]⊤ {\displaystyle \ X=\left[\ X_{1},X_{2},\ldots ,X_{m}\ \right]^{\top }\ } and Y=[ Y1,Y2,…,Yn ]⊤ {\displaystyle \ Y=\left[\ Y_{1},Y_{2},\ldots ,Y_{n}\ \right]^{\top }\ } be the i.i.d. sample vectors (for m≥n {\displaystyle \ m\geq n\ }) from Norm( μX, σX2 ) {\displaystyle \ {\mathsf {Norm}}\left(\ \mu _{\mathsf {X}},\ \sigma _{\mathsf {X}}^{2}\ \right)\ } and Norm( μY, σY2 ) {\displaystyle \ {\mathsf {Norm}}\left(\ \mu _{\mathsf {Y}},\ \sigma _{\mathsf {Y}}^{2}\ \right)\ } separately.

Let (P⊤)n×n {\displaystyle \ (P^{\top })_{n\times n}\ } be ann×n{\displaystyle n\times n} orthogonal matrix whose elements of the first row are all 1n  ,{\displaystyle \ {\tfrac {1}{\sqrt {n\ }}}\ ,} similarly, let (Q⊤)n×m {\displaystyle \ (Q^{\top })_{n\times m}\ } be the first n {\displaystyle \ n\ } rows of an m×m {\displaystyle \ m\times m\ } orthogonal matrix (whose elements of the first row are all 1m  {\displaystyle \ {\tfrac {1}{\sqrt {m\ }}}\ }).

Then Z≡ (Q⊤)n×m X m  −  (P⊤)n×n Y n  {\displaystyle \ Z\equiv {\frac {\ \left(Q^{\top }\right)_{n\times m}\ X\ }{\sqrt {m\ }}}\ -\ {\frac {\ \left(P^{\top }\right)_{n\times n}\ Y\ }{\sqrt {n\ }}}\ } is ann-dimensional normal random vector:

Z ∼ Norm( [ μX−μY, 0, 0, …, 0 ]⊤ , ( σX2 m+ σY2 n) In ) .{\displaystyle Z~\sim ~{\mathsf {Norm}}\left(\ \left[\ \mu _{\mathsf {X}}-\mu _{\mathsf {Y}},\ 0,\ 0,\ \ldots ,\ 0\ \right]^{\top }\ ,\ \left({\frac {\ \sigma _{\mathsf {X}}^{2}\ }{m}}+{\frac {\ \sigma _{\mathsf {Y}}^{2}\ }{n}}\right)\ I_{n}\ \right)~.}

From the above distribution we see that the first element of the vectorZ is

Z1=X¯−Y¯=1 m ∑i=1m Xi−1 n ∑j=1n Yj ,{\displaystyle Z_{1}={\bar {X}}-{\bar {Y}}={\frac {1}{\ m\ }}\sum _{i=1}^{m}\ X_{i}-{\frac {1}{\ n\ }}\sum _{j=1}^{n}\ Y_{j}\ ,}

hence the first element is distributed as

Z1−(μX−μY) ∼ Norm( 0,  σX2 m+ σY2 n ) ,{\displaystyle Z_{1}-\left(\mu _{\mathsf {X}}-\mu _{\mathsf {Y}}\right)~\sim ~{\mathsf {Norm}}\left(\ 0,\ {\frac {\ \sigma _{\mathsf {X}}^{2}\ }{m}}+{\frac {\ \sigma _{\mathsf {Y}}^{2}\ }{n}}\ \right)\ ,}

and the squares of the remaining elements ofZ arechi-squared distributed

 ∑i=2nZi2  n−1  ∼  χn−12  n−1 ×( σX2 m+ σY2 n){\displaystyle {\frac {\ \sum _{i=2}^{n}Z_{i}^{2}\ }{\ n-1\ }}~\sim ~{\frac {\ \chi _{n-1}^{2}\ }{\ n-1\ }}\times \left({\frac {\ \sigma _{\mathsf {X}}^{2}\ }{m}}+{\frac {\ \sigma _{\mathsf {Y}}^{2}\ }{n}}\right)}

and by construction of the orthogonal matriciesP andQ we have

Z1−(μX−μY)⊥∑i=2nZi2 ,{\displaystyle Z_{1}-\left(\mu _{\mathsf {X}}-\mu _{\mathsf {Y}}\right)\quad \perp \quad \sum _{i=2}^{n}Z_{i}^{2}\ ,}

soZ1, the first element ofZ, is statistically independent of the remaining elements by orthogonality.Finally, take for the test statistic

Te ≡  Z1−(μX−μY)  (∑i=2nZi2)/(n−1)   ∼ tn−1 .{\displaystyle T_{\mathsf {e}}~\equiv ~{\frac {\ Z_{1}-\left(\mu _{\mathsf {X}}-\mu _{\mathsf {Y}}\right)\ }{\ {\sqrt {\left(\sum _{i=2}^{n}Z_{i}^{2}\right)/\left(n-1\right)\ }}\ }}~\sim ~t_{n-1}~.}

Dependentt-test for paired samples

[edit]

This test is used when the samples are dependent; that is, when there is only one sample that has been tested twice (repeated measures) or when there are two samples that have been matched or "paired". This is an example of apaired difference test. Thet statistic is calculated as

t=X¯D−μ0sD/n,{\displaystyle t={\frac {{\bar {X}}_{D}-\mu _{0}}{s_{D}/{\sqrt {n}}}},}

whereX¯D{\displaystyle {\bar {X}}_{D}} andsD{\displaystyle s_{D}} are the average and standard deviation of the differences between all pairs. The pairs are e.g. either one person's pre-test and post-test scores or between-pairs of persons matched into meaningful groups (for instance, drawn from the same family or age group: see table). The constantμ0 is zero if we want to test whether the average of the difference is significantly different. The degree of freedom used isn − 1, wheren represents the number of pairs.

Example of matched pairs
PairNameAgeTest
1John35250
1Jane36340
2Jimmy22460
2Jessy21200
Example of repeated measures
NumberNameTest 1Test 2
1Mike35%67%
2Melanie50%46%
3Melissa90%86%
4Mitchell78%91%

Worked examples

[edit]
This articlemay need tosummarize its corresponding main article in better quality. Please help out toedit this article and make improvements to the summary. (Learn how and when to remove this message)

LetA1 denote a set obtained by drawing a random sample of six measurements:

A1={30.02, 29.99, 30.11, 29.97, 30.01, 29.99}{\displaystyle A_{1}=\{30.02,\ 29.99,\ 30.11,\ 29.97,\ 30.01,\ 29.99\}}

and letA2 denote a second set obtained similarly:

A2={29.89, 29.93, 29.72, 29.98, 30.02, 29.98}{\displaystyle A_{2}=\{29.89,\ 29.93,\ 29.72,\ 29.98,\ 30.02,\ 29.98\}}

These could be, for example, the weights of screws that were manufactured by two different machines.

We will carry out tests of the null hypothesis that themeans of the populations from which the two samples were taken are equal.

The difference between the two sample means, each denoted byXi, which appears in the numerator for all the two-sample testing approaches discussed above, is

X¯1−X¯2=0.095.{\displaystyle {\bar {X}}_{1}-{\bar {X}}_{2}=0.095.}

The samplestandard deviations for the two samples are approximately 0.05 and 0.11, respectively. For such small samples, a test of equality between the two population variances would not be very powerful. Since the sample sizes are equal, the two forms of the two-samplet-test will perform similarly in this example.

Unequal variances

[edit]

If the approach for unequal variances (discussed above) is followed, the results are

s12n1+s22n2≈0.04849{\displaystyle {\sqrt {{\frac {s_{1}^{2}}{n_{1}}}+{\frac {s_{2}^{2}}{n_{2}}}}}\approx 0.04849}

and the degrees of freedom

d.f.≈7.031.{\displaystyle {\text{d.f.}}\approx 7.031.}

The test statistic is approximately 1.959, which gives a two-tailed testp-value of 0.09077.

Equal variances

[edit]

If the approach for equal variances (discussed above) is followed, the results are

sp≈0.08399{\displaystyle s_{p}\approx 0.08399}

and the degrees of freedom

d.f.=10.{\displaystyle {\text{d.f.}}=10.}

The test statistic is approximately equal to 1.959, which gives a two-tailedp-value of 0.07857.

Related statistical tests

[edit]

Alternatives to thet-test for location problems

[edit]

Thet-test provides an exact test for the equality of the means of two i.i.d. normal populations with unknown, but equal, variances. (Welch'st-test is a nearly exact test for the case where the data are normal but the variances may differ.) For moderately large samples and a one tailed test, thet-test is relatively robust to moderate violations of the normality assumption.[27] In large enough samples, thet-test asymptotically approaches thez-test, and becomes robust even to large deviations from normality.[19]

If the data are substantially non-normal and the sample size is small, thet-test can give misleading results. SeeLocation test for Gaussian scale mixture distributions for some theory related to one particular family of non-normal distributions.

When the normality assumption does not hold, anon-parametric alternative to thet-test may have betterstatistical power. However, when data are non-normal with differing variances between groups, at-test may have bettertype-1 error control than some non-parametric alternatives.[28] Furthermore, non-parametric methods, such as theMann-Whitney U test discussed below, typically do not test for a difference of means, so should be used carefully if a difference of means is of primary scientific interest.[19] For example, Mann-Whitney U test will keep the type 1 error at the desired level alpha if both groups have the same distribution. It will also have power in detecting an alternative by which group B has the same distribution as A but after some shift by a constant (in which case there would indeed be a difference in the means of the two groups). However, there could be cases where group A and B will have different distributions but with the same means (such as two distributions, one with positive skewness and the other with a negative one, but shifted so to have the same means). In such cases, MW could have more than alpha level power in rejecting the Null hypothesis but attributing the interpretation of difference in means to such a result would be incorrect.

In the presence of anoutlier, thet-test is not robust. For example, for two independent samples when the data distributions are asymmetric (that is, the distributions areskewed) or the distributions have large tails, then the Wilcoxon rank-sum test (also known as theMann–WhitneyU test) can have three to four times higher power than thet-test.[27][29][30] The nonparametric counterpart to the paired samplest-test is theWilcoxon signed-rank test for paired samples. For a discussion on choosing between thet-test and nonparametric alternatives, see Lumley, et al. (2002).[19]

One-wayanalysis of variance (ANOVA) generalizes the two-samplet-test when the data belong to more than two groups.

A design which includes both paired observations and independent observations

[edit]

When both paired observations and independent observations are present in the two sample design, assuming data are missing completely at random (MCAR), the paired observations or independent observations may be discarded in order to proceed with the standard tests above. Alternatively making use of all of the available data, assuming normality and MCAR, the generalized partially overlapping samplest-test could be used.[31]

Multivariate testing

[edit]

A generalization of Student'st statistic, calledHotelling'st-squared statistic, allows for the testing of hypotheses on multiple (often correlated) measures within the same sample. For instance, a researcher might submit a number of subjects to a personality test consisting of multiple personality scales (e.g. theMinnesota Multiphasic Personality Inventory). Because measures of this type are usually positively correlated, it is not advisable to conduct separate univariatet-tests to test hypotheses, as these would neglect the covariance among measures and inflate the chance of falsely rejecting at least one hypothesis (Type I error). In this case a single multivariate test is preferable for hypothesis testing.Fisher's Method for combining multiple tests withalpha reduced for positive correlation among tests is one. Another is Hotelling'sT2 statistic follows aT2 distribution. However, in practice the distribution is rarely used, since tabulated values forT2 are hard to find. Usually,T2 is converted instead to anF statistic.

For a one-sample multivariate test, the hypothesis is that the mean vector (μ) is equal to a given vector (μ0). The test statistic isHotelling'st2:

t2=n(x¯−μ0)′S−1(x¯−μ0){\displaystyle t^{2}=n({\bar {\mathbf {x} }}-{{\boldsymbol {\mu }}_{0}})'{\mathbf {S} }^{-1}({\bar {\mathbf {x} }}-{{\boldsymbol {\mu }}_{0}})}

wheren is the sample size,x is the vector of column means andS is anm ×msample covariance matrix.

For a two-sample multivariate test, the hypothesis is that the mean vectors (μ1,μ2) of two samples are equal. The test statistic isHotelling's two-samplet2:

t2=n1n2n1+n2(x¯1−x¯2)′Spooled−1(x¯1−x¯2).{\displaystyle t^{2}={\frac {n_{1}n_{2}}{n_{1}+n_{2}}}\left({\bar {\mathbf {x} }}_{1}-{\bar {\mathbf {x} }}_{2}\right)'{\mathbf {S} _{\text{pooled}}}^{-1}\left({\bar {\mathbf {x} }}_{1}-{\bar {\mathbf {x} }}_{2}\right).}

The two-samplet-test is a special case of simple linear regression

[edit]
This sectiondoes notcite anysources. Please helpimprove this section byadding citations to reliable sources. Unsourced material may be challenged andremoved.(March 2025) (Learn how and when to remove this message)

The two-samplet-test is a special case of simplelinear regression as illustrated by the following example.

A clinical trial examines 6 patients given drug or placebo. Three (3) patients get 0 units of drug (the placebo group). Three (3) patients get 1 unit of drug (the active treatment group). At the end of treatment, the researchers measure the change from baseline in the number of words that each patient can recall in a memory test.

Scatter plot with six point. Three points on the left and are aligned vertically at the drug dose of 0 units. And the other three points on the right and are aligned vertically at the drug dose of 1 unit.

A table of the patients' word recall and drug dose values are shown below.

Patientdrug.doseword.recall
101
202
303
415
516
617

Data and code are given for the analysis using theR programming language with thet.test andlmfunctions for the t-test and linear regression. Here are the same (fictitious) data above generated in R.

>word.recall.data=data.frame(drug.dose=c(0,0,0,1,1,1),word.recall=c(1,2,3,5,6,7))

Perform thet-test. Notice that the assumption of equal variance,var.equal=T, is required to make the analysis exactly equivalent to simple linear regression.

>with(word.recall.data,t.test(word.recall~drug.dose,var.equal=T))

Running the R code gives the following results.

  • The mean word.recall in the 0 drug.dose group is 2.
  • The mean word.recall in the 1 drug.dose group is 6.
  • The difference between treatment groups in the mean word.recall is 6 – 2 = 4.
  • The difference in word.recall between drug doses is significant (p=0.00805).

Perform a linear regression of the same data. Calculations may be performed using the R functionlm() for a linear model.

>word.recall.data.lm=lm(word.recall~drug.dose,data=word.recall.data)>summary(word.recall.data.lm)

The linear regression provides a table of coefficients and p-values.

CoefficientEstimateStd. Errort valueP-value
Intercept20.57743.4640.02572
drug.dose40.81654.8990.000805

The table of coefficients gives the following results.

  • The estimate value of 2 for the intercept is the mean value of the word recall when the drug dose is 0.
  • The estimate value of 4 for the drug dose indicates that for a 1-unit change in drug dose (from 0 to 1) there is a 4-unit change in mean word recall (from 2 to 6). This is the slope of the line joining the two group means.
  • The p-value that the slope of 4 is different from 0 is p = 0.00805.

The coefficients for the linear regression specify the slope and intercept of the line that joins the two group means, as illustrated in the graph. The intercept is 2 and the slope is 4.

Regression lines

Compare the result from the linear regression to the result from thet-test.

  • From thet-test, the difference between the group means is 6-2=4.
  • From the regression, the slope is also 4 indicating that a 1-unit change in drug dose (from 0 to 1) gives a 4-unit change in mean word recall (from 2 to 6).
  • Thet-testp-value for the difference in means, and the regression p-value for the slope, are both 0.00805. The methods give identical results.

This example shows that, for the special case of a simple linear regression where there is a single x-variable that has values 0 and 1, thet-test gives the same results as the linear regression. The relationship can also be shown algebraically.

Recognizing this relationship between thet-test and linear regression facilitates the use of multiple linear regression and multi-wayanalysis of variance. These alternatives tot-tests allow for the inclusion of additionalexplanatory variables that are associated with the response. Including such additional explanatory variables using regression or anova reduces the otherwise unexplainedvariance, and commonly yields greaterpower to detect differences than do two-samplet-tests.

Software implementations

[edit]

Manyspreadsheet programs and statistics packages, such asQtiPlot,LibreOffice Calc,Microsoft Excel,SAS,SPSS,Stata,DAP,gretl,R,Python,PSPP,Wolfram Mathematica,MATLAB andMinitab, include implementations of Student'st-test.

Language/ProgramFunctionNotes
Microsoft Excel pre 2010TTEST(array1,array2,tails,type)See[1]
Microsoft Excel 2010 and laterT.TEST(array1,array2,tails,type)See[2]
Apple NumbersTTEST(sample-1-values, sample-2-values, tails, test-type)See[3]
LibreOffice CalcTTEST(Data1; Data2; Mode; Type)See[4]
Google SheetsTTEST(range1, range2, tails, type)See[5]
Pythonscipy.stats.ttest_ind(a,b,equal_var=True)See[6]
MATLABttest(data1, data2)See[7]
MathematicaTTest[{data1,data2}]See[8]
Rt.test(data1, data2, var.equal=TRUE)See[9]
SASPROC TTESTSee[10]
JavatTest(sample1, sample2)See[11]
JuliaEqualVarianceTTest(sample1, sample2)See[12]
Statattest data1 == data2See[13]

See also

[edit]

References

[edit]
  1. ^The Microbiome in Health and Disease. Academic Press. 2020-05-29. p. 397.ISBN 978-0-12-820001-8.
  2. ^Szabó, István (2003). "Systeme aus einer endlichen Anzahl starrer Körper".Einführung in die Technische Mechanik (in German). Springer Berlin Heidelberg. pp. 196–199.doi:10.1007/978-3-642-61925-0_16 (inactive 1 November 2024).ISBN 978-3-540-13293-6.{{cite book}}: CS1 maint: DOI inactive as of November 2024 (link)
  3. ^Schlyvitch, B. (October 1937). "Untersuchungen über den anastomotischen Kanal zwischen der Arteria coeliaca und mesenterica superior und damit in Zusammenhang stehende Fragen".Zeitschrift für Anatomie und Entwicklungsgeschichte (in German).107 (6):709–737.doi:10.1007/bf02118337.ISSN 0340-2061.S2CID 27311567.
  4. ^Helmert (1876)."Die Genauigkeit der Formel von Peters zur Berechnung des wahrscheinlichen Beobachtungsfehlers directer Beobachtungen gleicher Genauigkeit".Astronomische Nachrichten (in German).88 (8–9):113–131.Bibcode:1876AN.....88..113H.doi:10.1002/asna.18760880802.
  5. ^Lüroth, J. (1876)."Vergleichung von zwei Werthen des wahrscheinlichen Fehlers".Astronomische Nachrichten (in German).87 (14):209–220.Bibcode:1876AN.....87..209L.doi:10.1002/asna.18760871402.
  6. ^Pfanzagl, J. (1996). "Studies in the history of probability and statistics XLIV. A forerunner of thet-distribution".Biometrika.83 (4):891–898.doi:10.1093/biomet/83.4.891.MR 1766040.
  7. ^Sheynin, Oscar (1995). "Helmert's work in the theory of errors".Archive for History of Exact Sciences.49 (1):73–104.doi:10.1007/BF00374700.ISSN 0003-9519.S2CID 121241599.
  8. ^Pearson, Karl (1895)."X. Contributions to the mathematical theory of evolution.—II. Skew variation in homogeneous material".Philosophical Transactions of the Royal Society of London A.186:343–414.Bibcode:1895RSPTA.186..343P.doi:10.1098/rsta.1895.0010.
  9. ^abStudent (1908)."The Probable Error of a Mean"(PDF).Biometrika.6 (1):1–25.doi:10.1093/biomet/6.1.1.hdl:10338.dmlcz/143545. Retrieved24 July 2016.
  10. ^"T Table".
  11. ^Wendl, Michael C. (2016). "Pseudonymous fame".Science.351 (6280): 1406.doi:10.1126/science.351.6280.1406.PMID 27013722.
  12. ^Walpole, Ronald E. (2006).Probability & statistics for engineers & scientists. Myers, H. Raymond (7th ed.). New Delhi: Pearson.ISBN 81-7758-404-9.OCLC 818811849.
  13. ^Raju, T. N. (2005). "William Sealy Gosset and William A. Silverman: Two 'Students' of Science".Pediatrics.116 (3):732–735.doi:10.1542/peds.2005-1134.PMID 16140715.S2CID 32745754.
  14. ^Dodge, Yadolah (2008).The Concise Encyclopedia of Statistics. Springer Science & Business Media. pp. 234–235.ISBN 978-0-387-31742-7.
  15. ^Fadem, Barbara (2008).High-Yield Behavioral Science. High-Yield Series. Hagerstown, MD: Lippincott Williams & Wilkins.ISBN 9781451130300.
  16. ^Rice, John A. (2006).Mathematical Statistics and Data Analysis (3rd ed.). Duxbury Advanced.[ISBN missing]
  17. ^Weisstein, Eric."Student'st-Distribution".mathworld.wolfram.com.
  18. ^David, H. A.; Gunnink, Jason L. (1997). "The Pairedt Test Under Artificial Pairing".The American Statistician.51 (1):9–12.doi:10.2307/2684684.JSTOR 2684684.
  19. ^abcdLumley, Thomas;Diehr, Paula; Emerson, Scott; Chen, Lu (May 2002)."The Importance of the Normality Assumption in Large Public Health Data Sets".Annual Review of Public Health.23 (1):151–169.doi:10.1146/annurev.publhealth.23.100901.140546.ISSN 0163-7525.PMID 11910059.
  20. ^Markowski, Carol A.; Markowski, Edward P. (1990). "Conditions for the Effectiveness of a Preliminary Test of Variance".The American Statistician.44 (4):322–326.doi:10.2307/2684360.JSTOR 2684360.
  21. ^Guo, Beibei; Yuan, Ying (2017). "A comparative review of methods for comparing means using partially paired data".Statistical Methods in Medical Research.26 (3):1323–1340.doi:10.1177/0962280215577111.PMID 25834090.S2CID 46598415.
  22. ^Bland, Martin (1995).An Introduction to Medical Statistics. Oxford University Press. p. 168.ISBN 978-0-19-262428-4.
  23. ^"Central limit theorem and the normality assumption > Normality > Continuous distributions > Distribution > Statistical Reference Guide | Analyse-it® 6.15 documentation".analyse-it.com. Retrieved2024-05-17.
  24. ^DEMİR, Süleyman (2022-06-26)."Comparison of Normality Tests in Terms of Sample Sizes under Different Skewness and Kurtosis Coefficients".International Journal of Assessment Tools in Education.9 (2):397–409.doi:10.21449/ijate.1101295.ISSN 2148-7456.
  25. ^Minimum Detectable Difference for Two-Sample t-Test for Means. Equation and example adapted from Zar, 1984
  26. ^Wang, Chang; Jia, Jinzhu (2022). "Te Test: A New Non-asymptotic T-test for Behrens-Fisher Problems".arXiv:2210.16473 [math.ST].
  27. ^abSawilowsky, Shlomo S.; Blair, R. Clifford (1992). "A More Realistic Look at the Robustness and Type II Error Properties of thet Test to Departures From Population Normality".Psychological Bulletin.111 (2):352–360.doi:10.1037/0033-2909.111.2.352.
  28. ^Zimmerman, Donald W. (January 1998). "Invalidation of Parametric and Nonparametric Statistical Tests by Concurrent Violation of Two Assumptions".The Journal of Experimental Education.67 (1):55–68.doi:10.1080/00220979809598344.ISSN 0022-0973.
  29. ^Blair, R. Clifford; Higgins, James J. (1980). "A Comparison of the Power of Wilcoxon's Rank-Sum Statistic to That of Student'st Statistic Under Various Nonnormal Distributions".Journal of Educational Statistics.5 (4):309–335.doi:10.2307/1164905.JSTOR 1164905.
  30. ^Fay, Michael P.; Proschan, Michael A. (2010)."Wilcoxon–Mann–Whitney ort-test? On assumptions for hypothesis tests and multiple interpretations of decision rules".Statistics Surveys.4:1–39.doi:10.1214/09-SS051.PMC 2857732.PMID 20414472.
  31. ^Derrick, B; Toher, D; White, P (2017)."How to compare the means of two samples that include paired observations and independent observations: A companion to Derrick, Russ, Toher and White (2017)"(PDF).The Quantitative Methods for Psychology.13 (2):120–126.doi:10.20982/tqmp.13.2.p120.

Sources

[edit]

Further reading

[edit]
  • Boneau, C. Alan (1960). "The effects of violations of assumptions underlying thet test".Psychological Bulletin.57 (1):49–64.doi:10.1037/h0041412.PMID 13802482.
  • Edgell, Stephen E.; Noon, Sheila M. (1984). "Effect of violation of normality on thet test of the correlation coefficient".Psychological Bulletin.95 (3):576–583.doi:10.1037/0033-2909.95.3.576.

External links

[edit]
Wikiversity has learning resources aboutt-test
Wikisource has original text related to this article:
Continuous data
Center
Dispersion
Shape
Count data
Summary tables
Dependence
Graphics
Study design
Survey methodology
Controlled experiments
Adaptive designs
Observational studies
Statistical theory
Frequentist inference
Point estimation
Interval estimation
Testing hypotheses
Parametric tests
Specific tests
Goodness of fit
Rank statistics
Bayesian inference
Correlation
Regression analysis
Linear regression
Non-standard predictors
Generalized linear model
Partition of variance
Categorical
Multivariate
Time-series
General
Specific tests
Time domain
Frequency domain
Survival
Survival function
Hazard function
Test
Biostatistics
Engineering statistics
Social statistics
Spatial statistics
Retrieved from "https://en.wikipedia.org/w/index.php?title=Student%27s_t-test&oldid=1284704142"
Categories:
Hidden categories:
[8]ページ先頭
©2009-2025 Movatter.jp