It’s nice to think we’re all perfect researchers; that we're always running appropriate checks on our data, using the most up-to-date methods in statistics, and never overlooking problems in our analyses, but, in reality, that’s just not the case. A good stats app should address this by not only making stats easy for people, but helping them to become better researchers in the process; one of my key motivations for building statscloud.

I had a brief shriek of joy then when I discovered SIPS; the Society for Improving Psychological Sciences. After reading their mission statement, I was amazed at how closely aligned it was to the philosophy of statscloud (in improving training and research practices, and incentivising better scientific practices). Using this mission statement as a template, I was inspired to write this blog post and share how compatible statscloud is with these ideas.

Improving Training and Research Practices

Having been one of those people who struggled with stats, improving statistics training was probably my biggest motivation for building statscloud. The first step in achieving this was to design a brand new, modern user-interface; not a rehash of the same tired old design, but something new and fresh to support a new generation of psychologists. With apps generally heading away from desktops and towards touch devices too, it was the perfect opportunity to rethink the way stats software is designed.

One thing I’ve tried to get away from, for instance, are the analysis tick-boxes. When you’re a student learning statistics for the first time, and you’re just coming to grips with when to run a specific analysis, the last thing you need is a screen full of tick boxes badgering you like a needy child. OK, you’ve chosen your t-test, but what t-test? What about an effect size? Which one? Would you like a normality test? What about this one? How about a variance equality test? Do you want descriptives? Plots? Which ones? As they click away through these boxes, there’s no indication they’re selecting the things they actually need and no way of them interpreting whatever’s popped up on their screen. So, I thought that could use a change.

When you run a new analysis in statscloud, all you do is select the variables you want in your analysis, and it’ll take care of all the reliability checks for you. As standard, statscloud automatically calculates effect sizes, runs power calculations, and does all the appropriate reliability tests for the analysis you’ve selected. It alerts you whenever there are any issues with these (using a simple ‘traffic light’ system), and tucks all these tests away in a ‘reliability’ section below the main analysis which you can explore at your own leisure. If you’ve run an analysis that isn’t reliable though, these warnings hang around in the output until you’ve done something about them. Sorry about that ;-)

Improving institutional policies and norms to incentivise better scientific practices

Recently, psychologists have made great strides in helping us make research more open, but there’s still lots we can do to help get people on board with Open Science. It’s no secret that, while many are quite enthusiastic about changing the way we run and report research, others view it as a bit of an upheaval. However, I think the best way to influence people’s behaviour is to show them how easy it is to change it, and this is something statscloud was designed to do.

One way statscloud addresses this is by making it really easy to share open projects with people online. Thanks to the Open Science Framework, you can upload documents that outline what variables you’d like to measure in your experiment and what analyses you’re planning to run on your data. However, imagine just sharing a link to an open project with everything (the variables and analyses) all set up ready for data to be added... something like this.

These project templates are a great way of demonstrating to people what your experimental plan is before you’ve even collected any data, and sharing a project via a URL means anyone can open it up in their web browser on any device without downloading or installing any software. Everyone will be free to play around with the data, explore the analyses, and reshare it when they've finished. How about that?

Conducting empirical tests of the current state and reforms of scientific practices

It’s essential that a statistical app stays up to date with the best and most current empirical tests, and that it offers a range of different statistical methods too (e.g. frequentist, equivalence tests, Bayesian analysis, for instance). However, more important than that is that the app stays agnostic over which method of analysis you use and doesn’t push you towards any one statistical philosophy in particular. Because of this, statscloud is being developed to offer a whole range of methods and will allow you to choose the one that best fits your needs, and it won't judge you for it either.

Being a web app, it’s very easy for statscloud to ensure users stay up to date on the best practices in statistics too. Unlike a native (installed) app, you don’t need to worry about updating statscloud.app, or waiting around for a new version to be released with all the latest features. By keeping everything online, you can ensure you and your students are using the most up-to-date version of statscloud with all the most up-to-date tests.

One update made this year for instance was the way statscloud runs the Independent-samples one-way ANOVA. In response to the excellent article by Delacre et al (2019), the Welch’s F-test was added as an option to the One-way ANOVA (Independent) list and is now selected by default when running this test design. Now, when users select the standard F-test, they’re told that a newer and better test is available. I did the same last year with the Welch’s t-test following another excellent paper by Delacre et al (2017) and blog post by Daniel Lakens, and will continue to do the same for future tests. After all, stats is a fast moving area, so a stats app should be too.

Summary

These are just a few ways statscloud is aiming to support statistics and help people become better researchers. There are lots of other exciting things in the works to support good research practices, so I'll expand on these in a future blog post. Until then, make sure you stay up to date on all the latest updates by following statscloud on Twitter and Facebook and, if you haven't already, have a play around with statscloud at statscloud.app!