dstat: Descriptive statistics of quantitative analysis results

In quantchem: Quantitative chemical analysis: calibration and evaluation of results

Description

Performs comprehensive statistical evaluation of quantitative analysis results.

Usage

dstat(x, expected = median(unlist(x)), sort = TRUE, inverse.f = TRUE,
  na.rm = FALSE, conf.level = 0.95, alternative = c("two.sided", "less", "greater"), ansari = FALSE)

Arguments

x	a vector of results, of a dataframe with results to compare
expected	expected reference value
sort	if TRUE, the matrices are sorted by means, variances or p-values.
inverse.f	if F value in variance comparison is below 1, the inverse is taken (has no effect on p-value, but there are sometimes need to have such F
na.rm	logical: should NA values be removed?
conf.level	level for calculate confidence intervals
alternative	alternative for all tests performed.
ansari	due to reports of errors on some datasets, the ansari.test() on data is turned off by default since 0.12. you can turn it on by setting this variable to TRUE

Details

If argument is vector, several one-row matrices are produced (see below). If argument is a data.frame, there are also additional matrices with pairwise comparisons. The comparison of all groups by appropriate test are also calculated. This function prints its results with significance stars and returns a list invisibly.

Value

A list containing following matrices (if data is a vector, only 5 first are returned):

mean	mean, its confidence interval and t-test for expected value
median	median, its confidence interval and Wilcoxon test for expected value
var	variance, standard deviation, standard error and Dixon test for outlier
rsd	relative standard deviation, its confidence interval and Grubbs test for outlier
range	minimum and maximum value, range, IQR, MAD and Shapiro-Wilk test for normality
vartest	ratios of variances, their confidence intervals and F test with p-value
ttest	differences between means, their confidence intervals and t test with p-value
atest	nonparametric differences in scale, their confidence intervals and Ansari-Bradley test with p-value
atest	nonparametric differences in location, their confidence intervals and Wilcoxon test with p-value
anova	ANOVA between all data
kruskal	Kruskal-Wallis test (nonparametric equivalent for ANOVA)
bartlett	Bartlett test for homogeneity of all variances
fligner	Fligner-Killeen test for equal variances (nonparametric alternative to Bartlett)

Note

This function calculates always *both* parametric and nonparametric tests, and choosing a test to take into account should be also decision of analyst, based on the other tests results.

Author(s)

Lukasz Komsta

See Also

vstat

Examples

set.seed(1234)
a = data.frame(x=rnorm(10),y=runif(10),z=rt(10,1))
dstat(a,0)

Example output

Loading required package: MASS
Loading required package: outliers
------------------------------------------
If you use this package, please cite the recent paper containing description of this software:
Komsta, L. Chemometric and statistical evaluation of calibration curves in pharmaceutical analysis 
           - a short review on trends and recommendations. J. AOAC Int. 2012, 95, 3, 669-672.
------------------------------------------
Object: a 

Means:
      Mean    Lower    Upper       t   Pr(>t)   
x -0.38316 -1.09550  0.32919 -1.2168 0.254634   
y  0.41652  0.17867  0.65438  3.9614 0.003297 **
z  0.75803 -2.78798  4.30404  0.4836 0.640233   
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Medians:
    Median    Lower    Upper  W   Pr(>W)   
x -0.55554 -1.03413  0.35328 14 0.193359   
z -0.37616 -1.22575  6.41352 13 0.160156   
y  0.30965  0.17134  0.66815 55 0.001953 **
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Scale parameters and Dixon test:
  Variance       SD       SE      Q Pr(>Q)    
y  0.11055  0.33250  0.10514 0.0959 0.5881    
x  0.99159  0.99579  0.31490 0.3993 0.2187    
z 24.57168  4.95698  1.56754 0.8861 <2e-16 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

RSD Interval and Grubbs test:
       RSD    Lower    Upper      G   Pr(>G)    
x -259.890 -178.761 -474.458 1.9708   0.1323    
y   79.827   54.908  145.733 1.4982   0.5901    
z  653.932  449.797 1193.825 2.8107 5.14e-07 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Ranges and Shapiro test:
        Min       Max     Range       IQR       MAD      W    Pr(<W)    
y  0.039996  0.914658  0.874662  0.565389  0.355770 0.8845    0.1471    
x -2.345698  1.084441  3.430139  1.202414  1.100456 0.9551    0.7295    
z -2.084247 14.690528 16.774775  1.293172  0.580305 0.5112 4.921e-06 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Ratios of variances:
       Ratio    Lower    Upper        F    Pr(>F)    
y-z 222.2588 894.8126  55.2059 222.2588 3.522e-09 ***
x-z  24.7800  99.7642   6.1550  24.7800 5.276e-05 ***
x-y   8.9693   2.2278  36.1103   8.9693  0.003151 ** 
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Differences between means:
         Diff     Lower     Upper       t  Pr(>t)  
x-y -0.799679 -1.530520 -0.068838 -2.4088 0.03472 *
x-z -1.141185 -4.717340  2.434970 -0.7138 0.49215  
y-z -0.341506 -3.890660  3.207649 -0.2174 0.83272  
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Differences in locations:
        Diff    Lower    Upper  W   Pr(>W)   
y-z  0.81609  0.30877  1.90926 90 0.001505 **
x-y -0.77434 -1.41574  0.11838 26 0.075256 . 
x-z  0.26218 -0.74223  1.09406 54 0.795936   
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

ANOVA: F = 0.4008782 p-value = 0.67365 
Kruskal-Wallis test: chi-squared = 8.299355 p-value = 0.01577 

Bartlett test : K-Squared = 46.73898 p-value = 7.0919e-11 
Fligner-Killeen test: chi-squared = 3.446957 p-value = 0.17844 

quantchem documentation built on May 30, 2017, 5:28 a.m.