Parameters

Contingence Table

Download table

Parameters

Test Presentation

Download table

Test Construction

Test Presentation

Download table

Test Construction

Test Presentation

Download table

Model Construction

ANOVA

Adjusted means calculation

Contrast calculation

Model Construction

ANOVA

Adjusted means calculation

Contrast calculation

Survival curve creation

Cox Regression

Log Rank

Parameters

Graph

Parameters

Graph

Parameters

Graph

Statistical Analysis

TrialLytics

Introduction

The goal of this platform is to automate the statistical processes for studying the differences between multiple groups. Before explaining the tests performed, we will review the basic concepts needed to interpret these tests.

Prerequisites

  • Null Hypothesis (H0): The initial hypothesis to be tested. It is generally formulated as an equality or no significant difference between the groups.
  • Alternative Hypothesis (H1 or Ha): The hypothesis you aim to prove, typically formulated as a significant difference or inequality between the groups.
  • Significance Level (α): The probability threshold below which you reject the null hypothesis. It is often set at 0.05, meaning you accept a 5% risk of making a Type I error.
  • P-value: The probability of obtaining results as extreme as those observed if the null hypothesis is true. A low p-value (generally < α) suggests that you can reject the null hypothesis.

The Tests

Initially, the selected tests are:

  • ANOVA
  • Mixed Models
  • Kaplan-Meier Survival Curves
  • Cox Regression
  • Log Rank Test
  • Chi2
  • Student
  • Shapiro
  • Pearson

Linear Models

ANOVA

ANOVA is a parametric test used to determine if there is a significant difference between two or more groups for a given measurement. Group formation is done using categorical variables. The null hypothesis (H0) for this test is the equality of all groups (no significant difference), while the alternative hypothesis (H1) is that at least one group differs from the others. If the p-value is below our significance level, we reject the null hypothesis, and we can perform a Tukey test (Emmeans) to estimate the means of each group and determine which ones are different.

Example

In a study measuring PIE, the dependent variable would be PIE, and the explanatory variables could be the product and time (0H, 6H, etc.).

Emmeans

The Tukey test is used after rejecting the null hypothesis in an ANOVA. It helps determine which groups differ from each other.

Mixed Models

Mixed models combine ANOVA and the Tukey test, allowing us to detect if there are differences between multiple groups within a population and measure those differences if they exist. Unlike ANOVA, it relies on two types of effects:

  • Fixed Effects (the focus of the study)
  • Random Effects (not the focus of the study but can influence the results)

We assume that fixed effects react similarly across all individuals, while variations are captured by random effects.

Example

In a study measuring PIE, the dependent variable would be PIE. The fixed effects could be the product and time (0H, 6H, etc.), and the panelist would be a random effect. In a test studying the cellular response to different products, the dependent variable could be collagen production, the product concentration could be a fixed effect, and the cell line could be a random effect.

Kaplan-Meier Survival Curves

The Kaplan-Meier method is used to estimate survival probabilities over time. It allows for visualizing survival data and helps determine if different groups (e.g., treatment vs. control) exhibit different survival patterns over time.

This method is widely applied in clinical studies where the outcome of interest is time to event (such as time to death, disease recurrence, or other outcomes). The Kaplan-Meier curve provides an intuitive representation of survival data and can be compared across different groups.

Example

In a study measuring the longevity of lipstick wear, the event could be defined as the point when the lipstick wears off. Different curves can represent different lipstick products, allowing for comparison.

Survival

Cox Proportional Hazards Model

The Cox regression model, also known as the proportional hazards model, is used to assess the effect of multiple variables on survival time. Unlike Kaplan-Meier, which focuses on one group or treatment at a time, Cox regression allows for the inclusion of covariates and the examination of their influence on survival while adjusting for other factors.

This model assumes proportional hazards, meaning the effect of the covariates on the hazard (risk of event occurrence) is constant over time.

Example

In a study on the long-lasting effect of lipstick, factors such as product type, application technique, and environmental conditions could be included in the model to assess their effect on how long the lipstick stays on.

Log Rank Test

The Log Rank test is used to compare the survival distributions of two or more groups. It is a non-parametric test and can be used alongside Kaplan-Meier curves to determine whether there is a statistically significant difference between the survival curves of different groups.

The null hypothesis for the Log Rank test is that the survival curves are equal across the groups being compared.

Example

In a comparison of different lipstick brands, the Log Rank test could be used to determine if there is a significant difference in the longevity of wear across the different brands.

Other Statistical Tests

Chi-Square Test (Chi²)

The Chi-square test is a non-parametric test used to determine if there is a significant association between categorical variables. It compares observed and expected frequencies of events to assess whether distributions differ significantly across groups.

Example

In a study assessing customer preferences for different lipstick colors, the Chi-square test could determine if preferences vary significantly by age group.

Student’s t-Test

The Student’s t-test is a parametric test used to compare the means of two groups to see if they differ significantly. It assumes normally distributed data with equal variances across groups.

Example

In a cosmetic study comparing the average time it takes for different foundations to wear off, a t-test could assess if the mean wear times differ significantly between two specific products.

Shapiro-Wilk Test

The Shapiro-Wilk test is a normality test used to assess whether a sample is drawn from a normally distributed population. It is commonly used as a prerequisite check before applying parametric tests.

Example

In a study on foundation wear time, the Shapiro-Wilk test could confirm if wear time data follows a normal distribution, thus validating the use of parametric tests.

Pearson Correlation

Pearson correlation measures the strength and direction of the linear relationship between two continuous variables. It assumes both variables are normally distributed.

Example

In a study of customer satisfaction and repurchase rate for a skincare product, Pearson correlation could reveal if higher satisfaction is associated with increased repurchase rates.

Dimensionality Reduction

Principal Component Analysis (PCA)

PCA is a technique used to reduce the dimensionality of large datasets while retaining as much variability as possible. By transforming variables into a new set of uncorrelated components, PCA helps identify the main features driving data variation.

Example

In a study with many product characteristics (color, durability, texture), PCA could reduce the dimensions to a few principal components that summarize the main variation in product properties.

Multiple Correspondence Analysis (MCA)

MCA is similar to PCA but is designed for categorical data. It allows for the exploration of relationships between multiple categorical variables by reducing data dimensionality and creating easily interpretable visualizations.

Example

In a survey study, MCA could reveal relationships between customer demographics (age, gender) and preferences for specific product features.

t-Distributed Stochastic Neighbor Embedding (t-SNE)

t-SNE is a non-linear dimensionality reduction technique used for visualizing high-dimensional data in two or three dimensions. It excels at preserving local structures in the data, making it suitable for identifying clusters and patterns.

Example

In a dataset containing customer reviews of multiple products, t-SNE could help visualize groups of similar reviews, revealing patterns in customer feedback.

LEGAL NOTICE

In accordance with the provisions of Law No. 2004-575 of June 21, 2004, on confidence in the digital economy, users of the TrialLytics website are informed of the identity of the various parties involved in its creation and maintenance.

Hosting

The site is hosted by 1&1 / IONOS, located at 7 Place de la Gare - 57200 Sarreguemines, France.
Contact by phone or email: (+33) 9 70 80 89 11.

Publication Director

The Publication Director of the site is Aslane Mortreau.

Contact Us

By email: admin@mortreau.net

Personal Data

The processing of your personal data is governed by our Privacy Policy, available in the “Personal Data Protection Charter” section, in accordance with the General Data Protection Regulation 2016/679 of April 27, 2016 (“GDPR”).


Personal Data Protection Charter

Triallytics is committed to protecting user data in compliance with the General Data Protection Regulation (GDPR) 2016/679 of April 27, 2016. This charter explains how Triallytics processes and protects data when users interact with the platform.

1. Data Collection

Triallytics collects the following data to deliver its services:

  • User File: Users upload CSV/XLSX files for analysis. These files are processed solely to generate the requested results and are not stored.
  • Logs and User Location: Logs and country of access are recorded to analyze website traffic. No personal data is sold or shared.

2. Purpose of Data Processing

Data collected by Triallytics is processed for specific purposes:

  • Service Delivery: User files are used only to generate the results requested by the user.
  • Traffic Analysis: Logs and user location data help us understand daily traffic patterns and improve platform performance. No data is sold.

3. Data Storage and Retention

Triallytics does not store any user data permanently on the platform. Only daily visit counts are retained for analyzing site usage trends over time.

4. Data Sharing

Triallytics does not share any data with partners or third parties.

5. Data Security

No user data is stored on Triallytics. User files are processed temporarily to generate results, and no data remains on the platform after use.

6. User Rights

Since no user data is stored, Triallytics does not maintain any user information for access, rectification, or deletion requests.

7. Charter Updates

Any updates to this Personal Data Protection Charter will be communicated directly on the Triallytics platform. Users are encouraged to review this charter periodically to stay informed of any changes.


For further questions regarding this Personal Data Protection Charter, please contact us at admin@mortreau.net.