DataLab is a compact statistics package aimed at exploratory data analysis. Please visit the DataLab Web site for more information....



Statistical Tests

Command: Math -> Tests...

DataLab offers various statistical tests to be performed. The general procedure for performing any statistical test is as follows:

1) mark the data to be tested. Depending on the test to be performed it may be required to mark two blocks of data

2) select the test to be performed by clicking the corresponding button

3) reply to the questions asked by DataLab (if any)

4) the results are displayed in the protocol window

Please note, that the statistical tests as implemented in DataLab are performed on an "as is" basis. This means that the assumptions about the data are not checked (for example, the prerequisite that the variances are to be equal is not checked when performing a two-sample t-test). You have to check the validity of the underlying assumption at your own.

DataLab currently offers the following statistical tests:

Comparing Means and Medians
One-Sample t-Test Comparison of the mean of a sample with a predefined limit
Two-sample t-Test Comparison of the means of two samples with equal variances
Welch Test Comparison of the means of two samples with unequal variances
Mann-Whitney U-Test Comparison of the means of two samples of unknown distribution
t-Test for Paired Observations Comparison of the means of paired samples(1)
One-way ANOVA Simultaneous comparison of several means
Wilcoxon Signed Rank Test for Matched Pairs Comparison of the medians of two pairwise related samples
 
Comparing Variances
Chi-Square Test Comparison of the variance of a sample with a predifined limit
F Test Comparison of the variances of two samples
Levene's Test Simultaneous comparison of several variances  
 
Comparing Distributions(1)
Kolmogorov-Smirnov Test Testing against a standard normal distribution (mean = 0.0, std.dev. = 1.0)
Lilliefors Test Testing for normality (any mean or standard deviation)  
Shapiro-Wilk Test Testing for normality of small samples  
 
Checking Correlations
Significance of the Correlation Coefficient Comparison of the correlation coefficient with a predefined limit
 
Tests for Outliers
Variance/IQR-Test Three simples checks for outliers
Dean-Dixon Test Testing the minimum and the maximum of the data
Grubbs Test Test for outliers (assuming a normal distribution)



(1) The t-test for matched pairs requires the differences of the pairs to be normally distributed. In order to support the testing for normality of the paired differences the normality tests automatically calculate the results for the differences of two samples if they are pairwise related.


Last Update: 2012-Jul-25