My colleagues and I are currently looking into ODK as an option for a
large, multi-topic household survey. A few discussion threads have
mentioned in passing some problems (or anticipated problems)
implementing large surveys. That raises a few questions for me: some
theoretical, others practical.
In theory, is there an upper limit for the number of questions ODK can
handle? Are there any limitations inherent to the software? Or are
all limitations linked to hardware (e.g., processing power, memory,
etc.)?
In practice, how has the user community fared with large
questionnaires (note: by "large", I mean 300+ questions)? Is there
size threshold at which ODK becomes unstable or sluggish? Also, what
are the features, if any, test the practical limits of ODK (e.g.,
multiple languages+long questionnaires, nested loops, complex skip
instructions, etc.)?
First, if you have 300 questions, and each question takes 15 seconds to ask
and answer, the questionnaire will take 75 minutes to complete. So large
questionnaires are likely only useful if most of the questions are not
asked,
but are determined to be irrelevant due to earlier questioning. Many
taxonomy ("identify this animal") and diagnostic questionnaires are like
this.
Second, the underlying XForm processing engine in ODK Collect (javarosa)
eagerly constructs a full relevancy graph of what questions depend upon
answers
to earlier questions. That does take space, but, more importantly, it takes
time to construct, and, the larger the form, proportionately more time is
required. This leads to increasingly long start-up delays when a form is
loaded.
A better user experience would have the application lazily determine these
dependencies and to not construct this full relevancy graph since, for
larger
questionnaires, it can be expected that most questions will never be asked.
We are aware of this limitation but no solutions are imminent.
Mitch
···
On Tue, Nov 16, 2010 at 9:39 AM, Arthur wrote:
Dear ODKers,
My colleagues and I are currently looking into ODK as an option for a
large, multi-topic household survey. A few discussion threads have
mentioned in passing some problems (or anticipated problems)
implementing large surveys. That raises a few questions for me: some
theoretical, others practical.
In theory, is there an upper limit for the number of questions ODK can
handle? Are there any limitations inherent to the software? Or are
all limitations linked to hardware (e.g., processing power, memory,
etc.)?
In practice, how has the user community fared with large
questionnaires (note: by "large", I mean 300+ questions)? Is there
size threshold at which ODK becomes unstable or sluggish? Also, what
are the features, if any, test the practical limits of ODK (e.g.,
multiple languages+long questionnaires, nested loops, complex skip
instructions, etc.)?
number of questions and languages drive the load time. the sample
e-imci form has about 400 questions and 2 languages. it takes about 15
seconds to load on a nexus one. after loading, it's pretty snappy.
one way to get around the limitation is to have language specific or
smaller forms (ie, household form, adult form, child form). we've been
thinking about ways to link forms together, but as mitch noted, "no
solutions are imminent."
···
On Tue, Nov 16, 2010 at 11:10, Mitch Sundt wrote:
> There are two issues with large questionnaires.
>
> First, if you have 300 questions, and each question takes 15 seconds to ask
> and answer, the questionnaire will take 75 minutes to complete. So large
> questionnaires are likely only useful if most of the questions are not
> asked,
> but are determined to be irrelevant due to earlier questioning. Many
> taxonomy ("identify this animal") and diagnostic questionnaires are like
> this.
>
> Second, the underlying XForm processing engine in ODK Collect (javarosa)
> eagerly constructs a full relevancy graph of what questions depend upon
> answers
> to earlier questions. That does take space, but, more importantly, it takes
> time to construct, and, the larger the form, proportionately more time is
> required. This leads to increasingly long start-up delays when a form is
> loaded.
> A better user experience would have the application lazily determine these
> dependencies and to not construct this full relevancy graph since, for
> larger
> questionnaires, it can be expected that most questions will never be asked.
>
> We are aware of this limitation but no solutions are imminent.
>
> Mitch
>
>
> On Tue, Nov 16, 2010 at 9:39 AM, Arthur wrote:
>>
>> Dear ODKers,
>>
>> My colleagues and I are currently looking into ODK as an option for a
>> large, multi-topic household survey. A few discussion threads have
>> mentioned in passing some problems (or anticipated problems)
>> implementing large surveys. That raises a few questions for me: some
>> theoretical, others practical.
>>
>> In theory, is there an upper limit for the number of questions ODK can
>> handle? Are there any limitations inherent to the software? Or are
>> all limitations linked to hardware (e.g., processing power, memory,
>> etc.)?
>>
>> In practice, how has the user community fared with large
>> questionnaires (note: by "large", I mean 300+ questions)? Is there
>> size threshold at which ODK becomes unstable or sluggish? Also, what
>> are the features, if any, test the practical limits of ODK (e.g.,
>> multiple languages+long questionnaires, nested loops, complex skip
>> instructions, etc.)?
>>
>> Best,
>> Arthur
>>
>> --
>> Post: opendatakit@googlegroups.com
>> Unsubscribe: opendatakit+unsubscribe@googlegroups.com
>> Options: http://groups.google.com/group/opendatakit?hl=en
>
>
>
> --
> Mitch Sundt
> Software Engineer
> http://www.OpenDataKit.org
> University of Washington
> mitchellsundt@gmail.com
>
> --
> Post: opendatakit@googlegroups.com
> Unsubscribe: opendatakit+unsubscribe@googlegroups.com
> Options: http://groups.google.com/group/opendatakit?hl=en
>