Specification

The InVivoStat software package is made up of a number of analysis modules.

Summary Statistics

Summary statistics that can be calculated include the mean, standard deviation, SEM, min, max, median and coefficient of variation.

Single Measure Parametric Analysis

Analyses that can be carried out in this model include the t-test, ANOVA and ANCOVA approaches. Statistical results include ANOVA/ANCOVA table, residuals vs. predicted plots, normal probability plots, least square (predicted) means with confidence intervals, and post-hoc tests (unadjusted (LSD), Dunnett’s, Tukey, Holm, Hochberg, Hommel, Benjamini-Hochberg and Bonferroni.

Repeated Measures Parametric Analysis

Repeated measures analysis is carried out using a repeated measures mixed model approach, with the option to model the within-subject covariance as either compound symmetric, autoregressive or unstructured. Statistical results include overall effects table, residuals vs. predicted plots, normal probability plots, least square (predicted) means with confidence intervals and post-hoc unadjusted pairwise comparisons.

P-Value Adjustment

This module offers the user the ability to adjust p-values for multiplicity that have been calculated using other statistical software packages or other modules within InVivoStat. The p-value adjustment procedures available include Holm, Hochberg, Hommel, Benjamini-Hochberg and Bonferroni.

Non-Parametric Analysis

Non-parametric tests include Mann-Whitney tests (sometimes known as Wilcoxon tests), Kruskal-Wallis test, Behrens-Fisher all treatment comparisons and Steel’s all to one test.

Graphics

Graphical plots that can be created using this module include scatterplots (with best-fit lines where appropriate), observed means with standard errors (column or line version), histograms, box-plots and case profiles plots. All plots can be categorised by up to two factors and can be either overlaid or plotted separately in a lattice format. Note the properties of these plots (such as font, font size and font colour, plotting symbols and lines) can be changed within the Output Options window.

Power Analysis

The main purpose of this module is to provide the user with a power and sample size analysis plot. The input can either be an estimate of the mean and variance for the response of interest. Alternatively experimental data can be used with an option for the variance of the response to be adjusted for one factor (using a one-way ANOVA approach). The user has the option of considering percentage or actual change from control.

Dose-Response Analysis

This module fits a four-parameter logistic curve to dose-response data. Up to three of the parameters can be fixed. Quality control samples and unknown samples can also be included in the analysis. For each quality control, the relative error and coefficient of variation are calculated (for use in assay acceptance criteria assessment).

Alternatively the user can enter any non-linear equation of their choice. Start parameters for all unknown parameters that define the non-linear curve must also be defined.

Paired t-test/within-subject Analysis

This module allows the user to perform a paired t-test (when the within-subject ‘treatment’ factor has only two levels) or more generally a within-subject repeated measures analysis (when the within-subject ‘treatment’ factor has more than two levels). The module uses a repeated measures mixed model approach. The module is an extension of the Repeated Measures Parametric Analysis module for the case where there are no between–subject treatment factors.

Unpaired t-test Analysis

This module performs an unpaired t-test assuming either equal variance or unequal variance (Welch’s test).  The former is equivalent to the analysis performed in the Single Measures Parametric Analysis module if the analysis involves a single treatment factor that has only two levels.

 

Chi-squared and Fisher’s Exact test Analysis

This module allows the user to analyse categorical data. The data may be in list format or in the form of a contingency table. Analysis approaches include the Chi-squared test, Fisher’s Exact test and Barnard’s test.

 

Correlation Analysis

This module allows the user to correlate two or more continuous variables, categorized by up to four categorical factors. Correlation methods available include Pearson, Spearman and Kendall. The module also produces scatterplots, including best-fit lines, and overall matrix plots of all responses.

Survival Analysis

The Survival Analysis module allows the user to calculate the Kaplan-Meier non-parametric maximum likelihood estimates of the survival function, produce a plot of the survival curves and also perform a log-rank (or Mantel-Haenszel) test to compare the survival times.

 

Regression Analysis

The Regression Analysis module performs linear regression and multiple linear regression. The user can fit a model that includes continuous factors, multiple treatment (factorial) factors, other design (blocks) factors and a single covariate. The output includes statistics to assess the statistical model (R-squared, Cook’s distance plot and Leverage plot) as well as the other analysis assumption plots (the residuals vs. predicted plot and the normal probability plot).