6

Over the past 2 years myself and a colleague have built an online statistical analysis application using a mixture of silverlight, wcf and R. I (a c# programmer) wrote all the silverlight and wcf stuff whilst my colleague (a statistician) came up with the stats algorithms and wrote the R code. Now we think that this app is fairly unique - a rich gui online statistics application that is much more intuitive than all the other online stat apps that I've seen. But despite this we don't really know where to go with the project, mainly for the following reasons:

1) Its fairly complicated stuff - without the mix of programming and stats skills it would be difficult for anyone to "get into" the project and contribute.

2) We are stalled by a lack of a proper place to host the site. Currently it sits on the family windows 7 media centre, not exactly the best place to host it as it could interfere with the missus trying to watch Corrie/Friends/Oprah etc.

Soo, anyone got any ideas on how to move forward with this? I guess that my strength is programming not marketing so despite working hard at this for the past couple of years I feel that I've reached a dead end!

Also, does anyone know of any free windows hosting for open source projects? If I could find a proper place to put the app I might feel re-energised about the whole thing.

The source code is on codeplex at: http://silverstats.codeplex.com, whilst the app is currently hosted at http://silverstats.co.uk

Edit: to answer some of the points raised: I don't have any documentation so that is a valid point and something that I need to look at. However, I do believe that it is easy to use - try it yourself! It is drag and drop based and fairly intuitive (as long you have some stats knowledge). One point though: Its not aimed at statistians! Its aimed at people who need to do stats to do their job (which is why it is easy to use). Personally on an ease of use basis it beats the like of Rcmdr hands down! You are quite right that the R and stats geeks will continue to tap away at R and SAS and produce their scripts. So, to answer another question: Is it scriptable - of course not because that would defeat the ease of use concept, you might as well fire up R and tap away... Does it produce graphics: yep - output is nice and clean in html with some expalantory text for the people scared by stats. Is it validated, yes and no...the project is a spin off from our closed source (but free!) stats package called InVivoStat (google it) which was built with the same aim: to make statistics easy for non-statisticians, in this case scientists working in the life sciences. The R scripts that InVivoStat uses are fully GxP validated with evidence hosted on the InVivoStat website. This project (SilveR) uses the same scripts, so you could argue that a bit of revalidation is required as the interface has changed, but I am fully confident that the R scripts are correct.

Improve this question

5 Answers 5

Reset to default
4

To answer JBRWilkinson : yes, there is www.crossvalidated.com, a stackexchange site for statistics and well past beta. So that's a first channel to find both programmers and customers.

I agree with Lenny222 : you have to present a very clear case why your application is to be preferred, by showing :

  • it is easy to use
  • it is complete, i.e. all necessary statistical techniques are implemented.
  • it is correct, meaning that the results can be trusted
  • it is readible, meaning the output actually means something
  • it is able to produce custom graphics that explain the analysis in a straightforward way
  • it allows for scripting an analysis, so any analysis can be reproduced exactly

The latter point is extremely important, as you have to be able to repeat any published analysis on demand. I know it's not done like this for every publication, but at our university a statistical package will never be used if you can't script the analysis.

I hope I don't disappoint you too much, but you have to be realistic in this. R adepts are unlikely to jump on it as they like coding more and there are some very fine GUIs using R already on the "market" like JGR and Rcmdr. Statisticians using other programs (SAS, SPSS, Statistica, Stata, SPlus, ...) will probably keep on coding in the scripting language they like, and all the rest will just use the program they've been taught at university, or the one available on the servers of the college / company.

Maybe your best shot is showing the R community that your application can beat the other ones mentioned here and forms the bridge towards R coding for the students we have. If your application can do that, I'll be the first one to promote it here in statistical education. But you'll have to show it first, and I can tell you the competition is rather strong...

EDIT :

Given your further comments (and my look at the application), your best shot will be to provide an analysis guide, not only explaining the program but also explaining how to analyse different kinds of data with it, which statistics to use when, how to interprete results,... The application looks neat to me and can indeed be an extra value, but as the customers you aim for are no statisticians whatsoever, they need a lot more guidance. Make the guide visible on Google, and the rest will follow hopefully.

You could attempt a publication in a journal related to the field of research you're aiming at (as a letter to the editor or so maybe), but that's not going to be easy...

Improve this answer
8
  • I've edited the question to answer some of your points
    – Calanus
    Feb 3, 2011 at 20:25
  • @Calanus : don't take my comment as critics on your attempt, I'm just trying to be realistic here. I work as a statistician at the university of Ghent, and have a fair idea about what is used by whom. Non-statisticians will continue to use SPSS (or Statistica or so) until a statistician tells them to do something else. And - as you should know - a big part of this behaviour is coming from the peer review. I have had reviewers complain about using R for statistical analysis, as they didn't know it. How would they go about a free-to-download package?
    – Joris Meys
    Feb 3, 2011 at 20:48
  • @Calanus : with a thorough guide through the analysis (not only how to use the program, but which statistics to use with which data) you'll gain some public you want (I presume it's the experimental biology you look at). I wish you all the best, the app looks pretty neat to me.
    – Joris Meys
    Feb 3, 2011 at 20:52
  • It's ok I appreciate the construvtive criticism. You are right that non-statisticians will use whatever statisticians will tell them to use: In fact this is how InVivoStat was born was it was developed alongside a stats training course for life scientists which uses the application. SilveR is a spin off from this trying to reach a wider non-life science audience. However maybe you are right and this audience is impossible to reach without major statistical backing, although SilveR is free whereas SPSS and Statistica are not.
    – Calanus
    Feb 3, 2011 at 20:59
  • Peer review? No problem - a paper detailing the methodology behind InVivoStat is in prep, which also goes through a comparison with SPSS, Statistics, SAS, Unistat and others.
    – Calanus
    Feb 3, 2011 at 20:59
4

Do you have documentation? Not just the API ... tutorials, guides which make people think "wow, that's nice/shiny/easy/useful"? Do you have screenshots? Is your project easy to get up and running?

Regarding hosting: i think that should be easy to solve. Just rent a virtual server where you can do anything you'd like.

Maybe people's itch is not scratched hard enough by your project?

I also think that sometimes projects just fall between stools. I once wrote an Open Source vector graphic editor of Mac OS X. Many people wanted to use it, but there were almost nobody to contribute. Sometimes there is just no overlap between users and developers in the target audience.

Improve this answer
2
  • Fair comment regarding documentation - need to do something about that. However I do think that it will be hard to attract people to the project because you need to either be a good programmer and statistician, or have one of each!
    – Calanus
    Feb 3, 2011 at 20:43
  • Documentation is the thing that show that you just not only do the fun parts. The quality of the documentation tell even more.
    – user1249
    Jan 13, 2012 at 10:46
2

I'd suggest you to work on documentation. Write release notes, notes for developers, improve user manual, etc.

Improve this answer
2

Are you looking for more programmers to contribute to the code or are you really wanting people to start making use of using the app/library that you've developed?

Identify your target market and go find some early adopters. Is there a StackExchange community related to statistics? Is MathOverflow suitable?

Improve this answer
1
  • Both - I would like people to start using the application and also to contribute, it really needs both types for it to be a success
    – Calanus
    Feb 3, 2011 at 20:49
0

Sounds like you are looking for business advice rather programmers advice?

Codeplex is as good as any for hosting the code, you could try Sourceforge as well.

Improve this answer
4
  • 1
    I would be cautious regarding SF. I can not access the source code of some older projects of mine, after they got hacked recently. I prefer GitHub.
    – LennyProgrammers
    Feb 3, 2011 at 11:42
  • @Lenny222 GNU Savannah also had a problem with that. I'm self hosting all of my projects now.
    – Tim Post
    Feb 3, 2011 at 13:00
  • 2
    Ugh! Please, for the love of all that is binary, do not host your project on SourceForge if you have any shred of respect for your users or fellow developers. Even aside from the security issue, the interface is a mess and a half and it's way too much work to find stuff on there. I prefer Google Code; it's much cleaner and easier to navigate.
    – Mason Wheeler
    Feb 3, 2011 at 13:51
  • Not really looking for business advice: not really aiming to make any money out of it...
    – Calanus
    Feb 3, 2011 at 20:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.