Our goal is to make statistics open and accessible to everyone.
A stats package for everyone.
Statistics software is typically made by expert statisticians for other expert statisticians. Because of this, it assumes people open the software fully knowledgable on every test available, always choose the most appropriate test, and know how to interpret everything in the output.
In reality, that's just not the case. Many people open statistics software at different levels, including those who are just entering the field (students of psychology and other social sciences) and those who have stepped away from statistics for a while and require a refresh.
statscloud was created to address this issue. It aims to make statistics much more accessible to newcomers to statistics and also provide a more fluid user-experience for existing, professional statisticians.
- The Drug group had higher values for Memory Score than the Placebo group.
- The assumptions of this test were met.
- This test has a good effect size.
- This test has high power.
|Group||Descriptive statistics||Test statistics||Effect size|
|Outcome||Predictor||Group||n||Mean||Std. Dev.||df||t||p||Cohen's d||Hedge's g|
|Memory Score||Gender||Male||20||65.45||7.03||37.99||3.708||< .001||1.17||1.15|
The Male group had higher scores for Memory Score (M = 65.45, SD = 7.03) than the Female group (M = 57.15, SD = 7.13). An independent samples t-test (equal variances not assumed) showed this difference was statistically significant, t(37.99) = 3.71, p < .001, 95% Confidence Interval [4.53, 12.07], Cohen's ds = 1.17, observed power = 0.95. The common language effect size indicates that, if a pair of values were selected at random, there would be a 79.65% chance the Male value would be higher than the Female value.
Making stats available on any device, any time.
Designed for touch devices
It's not enough for an app simply to work on touch devices; a modern stats app should be designed for them. statscloud was built with a responsive design so that usability remains high on the most popular devices.
Lightweight and easy to install
Stats software is typically very large so it's not practical or possible to install on Chromebooks or touch devices. In contrast, statscloud has it's own lightweight statistics package built into it and the whole app can be installed on any device with a single click / tap.
Many individuals work with sensitive data and want to keep their data on their device. A web application that requires users to send their data over the internet would not be suitable, so statscloud is a Progressive Web App and was built to operate fully offline.
Helping people run reliable tests.
When running statistical analysis, it's important to make sure the test you have chosen is appropriate and all of its assumptions are met. To help with this, statscloud has some unique features to ensure your analyses are accurate and reliable.
Data cleaning is an important part of data analysis, but easy to overlook. To help you understand your data, statscloud provides a data summary for you which can alert you to any issues with your data before you start to run any analyses.
It's important for researchers to know when a test they've run is unreliable too, so statscloud produces a reliability report automatically with every analysis and suggests an alternative test if one is available.
(Cohen's ds = 1.17)
(Observed power = 0.95)
Teaching people how to code.
Instead, it shows people how to write code so they can do so themselves.
It's no secret that the best tool to use for statistical analysis is a programming language (e.g. R). Statistics packages with a point-and-click user-interface can distract people from learning how to use one.
Teaching a programming language at undergraduate level isn't always practical though because many students won't be interested in becoming computer programmers and may be put off degrees that require them to.
statscloud offers a comprise by offering a clear user-interface and showing the code for your project alongside it as you work. It annotates all of the code too so you can learn what is does you go along and migrate to a programming language whenever you're ready to.
Providing a range of analyses.
A summary of the current functionality, and the features that will be made available, are listed below.
|Independent-samples t-test||Support for both Welch's and the Student t-test.|
|One-way ANOVA (Independent)||Support for the standard F test, Welch's, and Brown-Forsythe test.|
|One-way ANOVA (Repeated)|
|ANOVA||Currently supports both Factorial and Repeated Measures ANOVAs. However, this is currently capped at 2 repeated-measures factors (with 10 levels). By the end of 2020, this will be adapted to support 10 factors. Sphericity corrections (e.g. Greenhouse-Geisser) will also be added shortly (by autumn 2020).|
|Linear Regression||Support for adding blocks (for hierarchical regression) will be arriving shortly in 2020.|
|Logistic Regression||Coming 2020|
|Mediation analysis||Coming 2020|
|Principal Components Analysis||Currently only supports an unrotated component matrix. However, other rotation methods (Varimax, Quartimax, Oblimin) will be added later in 2020.|
|McDonald's Omega||Coming 2020|
|Chi-Square||Support for chi-square tests with and without Yates continuity correction. Observed and expected values, standardised. unstandardised residuals are also available to view.|
|Cochran's Q test|
Equivilance (two one-sided) tests
Equivilance test versions of the analyses above will start to unroll in 2020 and will be complete in 2021.
Bayesian versions of the tests above will start to unroll in 2020 and will be complete in 2021.