You will find below the description of some utilities carried out under Excel in order to help you for several classical statistical tests. The solution with the R software is indicated in most of the utilities. They are free and can be downloaded, changed, copied and distributed freely. In order to avoid unintentional mistakes some pages are protected but without any password.

They are distributed without any guarantee

Except Box plots, they work with every Excel version since Excel 97
and also with Open Office (except some charts).

Non-parametric tests

Shapiro-Wilk's test
Mann & Whitney's test
Kruskal & Wallis test and pairwise comparisons
Wilcoxon's test
Friedman's test and pairwise comparisons
McNemar's test
Cochran's test
Dixon's test
Spearman's correlation test
Kendall's concordance test
Chi-square test
Non-parametric two-way analysis of variance

Parametric tests and tools

Two way Analysis of Variance
One way ANOVA for repeated measures
Two way ANOVA for repeated and independent measures
Three way ANOVA for repeated and independent measures
Bartlett's test
Hartley's Fmax test
Levene's test
Normality of a distribution
Computation of Confidence Intervals

Miscellaneous

Computation of the Power of a test
Choice of statistical’s tests
Control of the FDR for multiple hypothesis testing
Cronbach's alpha coefficient
Box Plots
Design of experiments
Statistical Tables

Choice of statistical’s tests

This Excel file presents a tree-like decision tool that facilitates the choice of a statistical’s test in most classical cases, according to the data form, the problematic and the variable type. The R software’s functions for each test are pointed out.

Download (44 Ko) [English version updated Nov. 23, 2009]

Statistical tables

About ten classical statistical tables are given: Spearman's Rho, Bravais-Pearson's r, Student's t, Fisher's, z, Chi-square, binomial law, Shapiro-Wilk's coefficients and table, Fmax, Durbin-Watson's DW. Instructions are given with each table

Download (345 Ko) [English version updated Oct. 13, 2008]

Two way Analysis of Variance

This tool performs a two-way analysis of variance for two independent factors with balanced sample size. A maximum of 7 levels each is permitted and the highest sample size is n = 30.
The validity conditions of the test are checked and a chart of the effects is given.

Download (68 Ko) [English version updated Feb. 2, 2010]

One way ANOVA for repeated measures

This tool performs a one-way analysis of variance for matched series. The data can reach 15 series and 100 individuals. The validity conditions of the test are checked and a chart of the effects is given.

Download (82 Ko) [English version updated Nov. 23, 2009]

Two way ANOVA for repeated and independent measures

This tool performs a two-way analysis of variance, one way for repeated measures and one way for independent samples. Series can be repeated until 15 times on 6 independent samples. The sample size cannot exceed 50 individuals and all samples must be of the same size.
The validity conditions of the test are checked and a chart of the effects is given.

Download (207 Ko) [English version updated Feb. 6, 2010]

Three way ANOVA for repeated and independent measures

This tool performs a two-way analysis of variance, one way for repeated measures and two ways for independent samples. Series can be repeated until 15 times on 5 independent samples. The sample size cannot exceed 30 individuals and all samples must be of the same size.
The validity conditions of the test are checked and a chart of the effects is given.

Download (141 Ko) [English version updated Feb. 6, 2010]

Non-parametric two-way analysis of variance

This tool performs a non-parametric two-way ANOVA for a factorial design with two levels in each independent group. Each factor may have until 6 categories and the sample size cannot exceed 30 individuals. The groups must be of the same size.

Download (93 Ko) [English version updated Jan. 29, 2009]
Bartlett's test

This tool performs the Bartlett's test of heterogeneity of variances of several independent groups. Until 50 variances can be compared. Independent samples can be of various sizes but distributions must be normal.

Download (56 Ko) [English version updated Apr. 30, 2010]

Box plots

Box plots are drawn for until 15 different groups.
Three different solutions are available depending on the specification of the tips of the whiskers.
This tool works only under Excel 2003.

Download (451 Ko) [English version updated Dec. 24, 2009]

Chi-square test

The Chi-square statistic, its significance and the Cramer's contingency coefficient are computed for contingency tables of usual sizes: max = 50 columns and 300 rows. The Yates correction is included.
The validity conditions of the test are indicated and checked.

Download (534 Ko) [English version updated May. 11, 2010]

Cochran's Q test

The Cochran's Q statistic and its significance are computed to compare more than two related series in a binomial scale. The dichotomous data are in a table of 0 and 1, the rows being the individuals and the columns being the number of groups or conditions.
The tool is limited to 15 series (columns) and 100 individuals (rows).

Download (29 Ko) [English version updated Dec. 4, 2009]
Cronbach's alpha coefficient

This tool computes the Cronbach's alpha coefficient to measure how well a set of items (or variables) measures a single unidimensional latent construct. Until 1,000 individuals and 50 items can be used.

Download (864 Ko) [English version updated Sep. 7, 2009]

Dixon's test

The Dixon's Q test is performed for identification and rejection of outliers. The sample size must be between 3 and 30.

Download (199 Ko) [English version updated Dec. 10, 2009]

Control of the FDR for multiple-hypothesis testing

This tool allows the control of the "False Discovery Rate" when numerous p-values are computed in a same experiment. It ensures to know what p-values really allow the rejection of the null hypothesis. Dependent and independent p-values are distinguished. Until 500 pt-values can be tested.
The solution to perform the same computation with the R software is indicated, as well as the way to compute the q-value.

Download (197 Ko) [English version updated may 29, 2009]

Hartley's Fmax test

This tool performs the Hartley's test, also called 'Fmax test'. The heterogeneity of variances of several independent groups is tested. Samples must have the same size and their distributions must be normal. Until 12 different samples can be compared.

Download (38 Ko) [English version updated Oct. 13, 2008]

Friedman's test and pairwise comparisons

This tool carries out the non-parametric Friedman's test for the comparison of more than two related samples. Non-parametric pairwise comparisons are performed, taking the number of groups into account.
It computes also the Kendall's coefficient of concordance.

Download (300 Ko) [English version updated Mar. 27, 2010]
Computation of Confidence Intervals

This tool computes Confidence Intervals of a mean, a proportion or a linear correlation.

Download (112 Ko) [English version updated Dec. 23, 2009]

Kendall's coefficient of concordance

This document does not process any computation but proposes solutions in connection with other tools for the computation of the Kendall's coefficient of concordance W which measures the concordance between several judges or between k variables.

Download (53 Ko) [English version updated Mar. 27, 2009]

Kruskal & Wallis test and pairwise comparisons

This tool carries out the non-parametric Kruskal & Wallis test for the comparison of more than two independent samples (until 10 groups with a maximum sample size of 50 individuals). Non-parametric pairwise comparisons are performed, taking the number of groups into account.

Download (215 Ko) [English version updated Jan. 27, 2010]

Levene's test

This tool performs the Levene's test of heterogeneity of variances of several independent groups. Until 12 groups can be compared with a maximum size of 300 individuals per sample.
Two solutions are available: the Levene's test (1960) using the differences with the means of the groups and the Brown & Forthyse's test (1974) using the differences with the medians of the groups.

Download (238 Ko) [English version updated Dec. 18, 2009]

McNemar's test

This tool performs the McNemar's change test for two related samples in nominal scale. The exact test is performed

Download (37 Ko) [English version updated Oct. 13, 2008]

Mann & Whitney's test

This tool allows the non-parametric comparison of two independent samples. Sample size must be under 30. Data can be either quantitative or in ordinal scale.

Download (74 Ko) [English version updated Dec. 10, 2009]

Normality of a distribution

Several normality indicators are computed by this tool: number of outliers, difference between mean and median, kurtosis and asymmetry coefficients, etc. No normality test as such is performed.

Download (227 Ko) [English version updated Oct. 13, 2008]

Design of experiments

Theses files allow the analysis of data from several experimental designs. The computation of minimum designs is allowed. Charts of effects and interactions are given and the ANOVA test is performed.

Download (334 Ko) [English version updated Jul. 12, 2010]

Computation of the Power of a test

This tool allows the computation of the power of a comparison test for two independent or related samples, for the comparison of two proportions, for a correlation test, for a one-way analysis of variance with 3, 4 or 5 independent groups.
It allows also the statistical comparison of two independent correlations.
The one-way ANOVA for 3, 4 or 5 independent groups is also performed, checking the validity conditions (homogeneity of variances).
Instructions and a numerical example are provided on the first sheets.

Download (167 Ko) [English version updated Jul. 16, 2010]
Shapiro-Wilk's test

This tool performs the Shapiro-wilk normality test for samples from 3 to 500 individuals. Basic statistics and some normality plots are also given. The template was build by Mohammad Ovais (Xepa-Soul Pattinson, Malaysia) and translated into French by AnaStat. Only plots are not available under Open Office.
A method to perform the test with the R software is indicated in the last sheet.

Download (58 Ko) [English version updated May. 3, 2010]

Spearman's correlation test

This tool computes the Spearman's Rho correlation coefficient and its significativity. It is a non-parametric test using the rows of the values. Sample size must be between 6 and 30.

Download (118 Ko) [English version updated Dec. 8, 2009]

Wilcoxon's test

This tool allows the non-parametric comparison of two related samples, with
5 < n < 30. Data must be in a quantitative scale of measurement.

Download (31 Ko) [English version updated Jan. 17, 2010]