One of the standout features of statscloud is that it helps users avoid making mistakes when running tests; it warns you if you've violated the assumptions of a test, suggests alternatives if you have, and lays out all this information for you in an interface you can interact with.

There may, however, be times when you don't want this functionality. If you're introducing students to a new type of analysis, you may want them to learn about the basics of the test before learning when it's appropriate to run it. You may also not want them to assume a test is reliable just because the app tells them it is; you may want them to decide this for themselves. Above all, you may just not want the app to make decisions about how reliable your analysis is or badger about you it when it does.

Well, we've got good news for you.

While statscloud has been in beta, we've collected some valuable information about what you'd like to see in the final release of the app and what could be done to improve it even further. I'm really pleased to tell you that we've recently made some changes to the app that take on board your comments and make the assumption checking system far more flexible and much more powerful.

Here's what we've done:

Opt-in rather than opt-out

For new users, automatic assumption checking system is now switched off by default. When you open up the app for the first time, you'll be greeted with a message asking if you'd like to turn automatic assumption checking on. When you do, the app will work exactly as it did before; the assumptions will be checked automatically and you'll be alerted if statscloud finds any issues with your tests. This setting only needs to be set once (your device will remember your preference next time you start the app) but you can toggle automatic assumption testing on and off any time by clicking the 'settings' icon in the analyses 'Reliability' section.

If you're already using the app with these settings and you're finding the assumption checks and warnings useful, don't worry; your preferences have already been saved so nothing will change for you! If you're ready to turn them off though, you can do that right now...

Decide for yourself

So, what happens if you don't have automatic assumption checking switched on? Well, the app will continue to provide information on all the assumptions of a test but, by default, it won't make any decisions about whether the assumptions have been met (they won't be lit in green, amber, or red). Instead, you're encouraged to open up the assumption pop-ups and decide this for yourself based on the information provided.

Involving the user when deciding if a test has met its assumptions allows to take a big step forward in assessing the reliability of tests. While there is clear, objective criteria for deciding whether certain assumptions have been met (e.g., a p value from Levene's test above/below 0.05 to check for variance equality), the assumptions of other analyses aren't quite as straightforward. Take, for instance, a regression analysis. One of the key assumptions of a linear regression is homoscedasticity; that the residuals in your regression model are equally varied. It's trickier to test an assumption like this using an objective test or formula so, in this situation, it's sometimes best to visualise the data instead (e.g., a scatter plot of the residuals and fitted values) and use this to make a judgement on whether the assumption has been met.

The recent modifications allow you to do exactly this. As always, statscloud provides an explanation of what the assumptions of the test are, but this time gives you the opportunity to log whether you think the assumption has been met yourself. At the bottom of the pop-up box, where statscloud asks you whether you think the assumption has been met, you can choose to respond "yes", "no", or say it is "unclear". If you choose to say an assumption has not been met, a red flag will then appear next to your analysis. If you're unsure, an amber flag will appear. If you say everything looks good for all your assumptions, you'll get a green flag.

This is a great way of encouraging users to think carefully about the issues related to an analysis and interact with them. By moving beyond objective assumption checking, we can now offer a new range of assumption tests that rely more on user input and guide you to make the important decisions for yourself. Of course, we can continue to add more assumption checks for tests when they're appropriate too. So, if there's something you'd like to see, do get in touch.

Power

Ah, 'observed power'. Everyone's favourite topic...

Many will be aware that post-hoc calculations of power have limited use in statistics, or, to be less polite, are essentially useless. To help users understand this, we've added a quick warning to the power analysis pop-up (see below):

The truth is, the power system in statscloud is part of a much wider power analysis system we're hoping to implement in the future. Here's a bit of background on this and what it will look like:

For some time now, users have been able to generate a template of an analysis; that is, you can set up the variables you would like to measure in an analysis before you've actually collected any data. When you have, and you've populated the spreadsheet, you can then 'run' this analysis and see the results.

This has all sorts of uses. Some lecturers find it useful to set up a template for an analysis and disseminate this link to students so all they have to do is enter data and then run the predefined analysis. Setting up templates for analyses is also good practice as it encourages the kind of behaviour used when filing preregistrations.

One of the other advantages of an analysis template is that it offers a prime opportunity to run an a-priori power calculation. So, before you've started your data collection, you can tell statscloud what analysis you would like to do and set up a power analysis as part of this template. Here, statscloud will then tell you how many subjects you need to test, and it will keep track of this throughout the data entry process, warning you if you've run the test too early and informing you when you've tested enough participants to meet the requirements of the power analysis.

When an analysis has been run, you'll have the opportunity to revisit the calculations in the current 'Power' pop-up and switch between an a-priori result and (should you want to) observed power.

More customisation

Maybe these changes are still not enough for you. Maybe you still don't want the app to decide what assumptions should be attached to each test, or don't even want your students to have the option to turn on automatic assumption testing. You may also not want power calculations to show up at all. Well, there's more good news:

We've been busy developing statscloud Enterprise; a version of the app that you can host yourself and have full control of. This will come with an administrator's dashboard that will allow you to customise what you'd like the app to look like and set up default options for your users. This will include the option to turn off the automatic assumption checking in the reliability section, customising which assumptions appear, or removing this system alogether.

We'll share more updates on when this version of statscloud will become available, but keep your eye on the Products page, and on social media (e.g., Twitter and Facebook) for updates on when this is ready.

Final thoughts

With statscloud, my goal has always been to make a statistics app that is much more intuitive and accessible for new users. Crucially, I've always believed that running statistical analysis is a reactive and dynamic process; that you need to be able to visualise and interact with your data as much as possible to understand exactly what it's telling you.

Our new approach to statistics allows users to do exactly this. With our modern, touch-friendly and cross-platform user-interface, we've made it possible for users to interact with the results of an analysis right inside the output, and have context provided every step of the way. Innovations like this continue to make statscloud truly unique, and help establish it as the new standard in statistics software design.