Performs a one-sample, two-sample, or a Welch modified two-sample
t-test based on user supplied summary information. Output is identical to that
produced with t.test.
tsum.test(mean.x, s.x = NULL, n.x = NULL, mean.y = NULL, s.y = NULL,
n.y = NULL, alternative = "two.sided", mu = 0, var.equal = FALSE,
conf.level = 0.95)a single number representing the sample mean of x
a single number representing the sample standard deviation for x
a single number representing the sample size for x
a single number representing the sample mean of y
a single number representing the sample standard deviation for y
a single number representing the sample size for y
is a character string, one of "greater", "less" or
"two.sided", or just the initial letter of each, indicating the specification
of the alternative hypothesis. For one-sample tests, alternative refers to the true
mean of the parent population in relation to the hypothesized value mu.
For the standard two-sample tests, alternative refers to the difference between
the true population mean for x and that for y, in relation to mu.
For the one-sample and paired t-tests, alternative refers to the true mean of the
parent population in relation to the hypothesized value mu. For the standard
and Welch modified two-sample t-tests, alternative refers to the difference between
the true population mean for x and that for y, in relation to mu.
For the one-sample t-tests, alternative refers to the true mean of the parent population
in relation to the hypothesized value mu. For the standard and Welch modified
two-sample t-tests, alternative refers to the difference between the true population
mean for x and that for y, in relation to mu.
is a single number representing the value of the mean or difference in means specified by the null hypothesis.
logical flag: if TRUE, the variances of the parent populations
of x and y are assumed equal. Argument var.equal should be supplied
only for the two-sample tests.
is the confidence level for the returned confidence interval; it must lie between zero and one.
A list of class htest, containing the following components:
the t-statistic, with names attribute "t"
is the degrees of freedom of the t-distribution
associated with statistic.
Component parameters has names attribute "df".
the p-value for the test.
is a confidence interval (vector of length 2)
for the true mean or difference in means. The confidence level
is recorded in the attribute conf.level. When alternative
is not "two.sided", the confidence interval will be half-infinite,
to reflect the interpretation of a confidence interval as the set of all
values k for which one would not reject the null hypothesis that
the true mean or difference in means is k . Here infinity will be
represented by Inf.
vector of length 1 or 2, giving the sample mean(s)
or mean of differences; these estimate the corresponding population
parameters. Component estimate has a names attribute describing its elements.
the value of the mean or difference in means specified by
the null hypothesis. This equals the input argument mu. Component
null.value has a names attribute describing its elements.
records the value of the input argument alternative:
"greater" , "less" or "two.sided".
a character string (vector of length 1) containing the names x and y for the two summarized samples.
For the one-sample t-test, the null hypothesis is that the mean of
the population from which x is drawn is mu. For the standard and Welch modified
two-sample t-tests, the null hypothesis is that the population mean for x less that for
y is mu.
The alternative hypothesis in each case indicates the direction of divergence of the population
mean for x (or difference of means for x and y) from mu
(i.e., "greater", "less", or "two.sided").
The assumption of equal population variances is central to the standard two-sample t-test. This test can be misleading when population variances are not equal, as the null distribution of the test statistic is no longer a t-distribution. If the assumption of equal variances is doubtful with respect to a particular dataset, the Welch modification of the t-test should be used.
The t-test and the associated confidence interval are quite robust with respect to level toward heavy-tailed non-Gaussian distributions (e.g., data with outliers). However, the t-test is non-robust with respect to power, and the confidence interval is non-robust with respect to average length, toward these same types of distributions.
For each of the above tests, an expression for the
related confidence interval (returned component conf.int) can be obtained in the usual
way by inverting the expression for the test statistic. Note that, as explained
under the description of conf.int, the confidence interval will be half-infinite when
alternative is not "two.sided" ; infinity will be represented by Inf.
If y is NULL, a one-sample t-test is
carried out with x. If y is not NULL, either a standard or
Welch modified two-sample t-test is performed, depending on whether var.equal is TRUE
or FALSE.
Kitchens, L.J. (2003). Basic Statistics and Data Analysis. Duxbury.
Hogg, R. V. and Craig, A. T. (1970). Introduction to Mathematical Statistics, 3rd ed. Toronto, Canada: Macmillan.
Mood, A. M., Graybill, F. A. and Boes, D. C. (1974). Introduction to the Theory of Statistics, 3rd ed. New York: McGraw-Hill.
Snedecor, G. W. and Cochran, W. G. (1980). Statistical Methods, 7th ed. Ames, Iowa: Iowa State University Press.
# NOT RUN {
round(tsum.test(mean.x=53/15, mean.y=77/11, s.x=sqrt((222-15*(53/15)^2)/14),
s.y=sqrt((560-11*(77/11)^2)/10), n.x=15, n.y=11, var.equal= TRUE)$conf, 2)
# Example 8.13 from PASWR
tsum.test(mean.x=4, s.x=2.89, n.x=25, mu=2.5)
# Example 9.8 from PASWR
# }
Run the code above in your browser using DataLab