[Qgis-developer] Kriging interpolation functionality in QGIS?

[Stefan Keller]

Hi,

I'm looking for a Kriging interpolation from an irregular point data
set to raster data in QGIS without dependencies, if possible. There's
1. Interpolation plugin (C++) [1]
2. „Raster Interpolation“ plugin von Denis Rouzaud (Python) [2]

I know that GRASS and SAGA implemented Kriging [3].
But these are are dependencies I'd like to avoid in order to make QGIS
even more easier to use.
AFAIK even Denis' "Raster Interpolation" plugin requires Scipy
separately installed.

So eventually, I'm considering to extend "Raster Interpolation" Python
plugin (or [1])
Any suggestions or activities regarding this?

:Stefan

[1] http://docs.qgis.org/2.8/de/docs/user_manual/plugins/plugins_interpolation.html
[2] http://3nids.github.io/rasterinterpolation/
[3] http://gis.stackexchange.com/questions/1041/open-source-methods-for-kriging

[Barry Rowlingson]

Post by Stefan Keller
Hi,
I'm looking for a Kriging interpolation from an irregular point data
set to raster data in QGIS without dependencies, if possible. There's
1. Interpolation plugin (C++) [1]
2. „Raster Interpolation“ plugin von Denis Rouzaud (Python) [2]
I know that GRASS and SAGA implemented Kriging [3].
But these are are dependencies I'd like to avoid in order to make QGIS
even more easier to use.
AFAIK even Denis' "Raster Interpolation" plugin requires Scipy
separately installed.
So eventually, I'm considering to extend "Raster Interpolation" Python
plugin (or [1])
Any suggestions or activities regarding this?

You want Kriging in QGIS? You now have to implement a full
statistical analysis system in QGIS. Kriging is hard, or even
impossible to do properly as an automatic points-to-raster process
without statistical diagnostics. You'll need to plot your variogram.
You'll need to fit your variogram. You may need to transform your
variables, and transform results back (and the back-transform of
Kriging standard-errors is not trivial).

The simplest points-to-raster Kriging system I've seen is autoKrige
from the R automap package. Even so this has about 7 parameters and
options to choose from. Plus working in R makes it easy to plot
diagnostics and histograms and do transforms and testing....

If you want to try all that, there's a few python kriging
implementations that you could use, or this blog poster shows how to
do it from scratch with basic matrix ops:

http://connor-johnson.com/2014/03/20/simple-kriging-in-python/

Barry

[Stefan Keller]

Hi Barry

Many thanks for your explanations and hints.
So from a pragmatic point-of-view ("80/20 pareto rule"):
Do you think Inverse Distance Weighting (IDW) would do the job as well,
since Kriging has so many parameters to fiddle around and to understand?

:Stefan

Post by Barry Rowlingson

Post by Stefan Keller
Hi,
I'm looking for a Kriging interpolation from an irregular point data
set to raster data in QGIS without dependencies, if possible. There's
1. Interpolation plugin (C++) [1]
2. „Raster Interpolation“ plugin von Denis Rouzaud (Python) [2]
I know that GRASS and SAGA implemented Kriging [3].
But these are are dependencies I'd like to avoid in order to make QGIS
even more easier to use.
AFAIK even Denis' "Raster Interpolation" plugin requires Scipy
separately installed.
So eventually, I'm considering to extend "Raster Interpolation" Python
plugin (or [1])
Any suggestions or activities regarding this?

You want Kriging in QGIS? You now have to implement a full
statistical analysis system in QGIS. Kriging is hard, or even
impossible to do properly as an automatic points-to-raster process
without statistical diagnostics. You'll need to plot your variogram.
You'll need to fit your variogram. You may need to transform your
variables, and transform results back (and the back-transform of
Kriging standard-errors is not trivial).
The simplest points-to-raster Kriging system I've seen is autoKrige
from the R automap package. Even so this has about 7 parameters and
options to choose from. Plus working in R makes it easy to plot
diagnostics and histograms and do transforms and testing....
If you want to try all that, there's a few python kriging
implementations that you could use, or this blog poster shows how to
http://connor-johnson.com/2014/03/20/simple-kriging-in-python/
Barry

[Barry Rowlingson]

Post by Stefan Keller
Hi Barry
Many thanks for your explanations and hints.
Do you think Inverse Distance Weighting (IDW) would do the job as well,
since Kriging has so many parameters to fiddle around and to understand?

Depends on what "the job" is. To get an impression of the overall
trend of a set of samples - IDW is probably fine. But to get a
*principled* (ie based on a statistical model) set of estimates over a
grid *with honest estimates of uncertainty* so you can answer
probabilistic questions (like "what's the chance that the soil over
there will contain 1ug gold/tonne?") you need something like kriging.

Barry

[Sjur Kolberg]

I'd prefer Kriging to IDW also for simple jobs.

Kriging is (at least) four operations:
0: Know the assumptions and check your data, transform if necessary and possible.
1: Estimate your empirical semivariogram (describing how difference increases with distance).
2: Select a parametric semivariogram model, and adjust parameters to fit the point cloud from 1)
3: Use the parametric model to calculate a weight matrix between each target location (grid point) and each observation point.

Loosely speaking, the three first operations define the core of Kriging. The last operation answers the same question as IDW, both providing the coefficients (weights) in a linear combination of the input data.

With Gaussian data, spatial stationarity, well-spread data points and a properly selected/tuned semivariogram model, Kriging can be shown to be 'Best linear unbiased estimator' (BLUE), and provides mathematically correct estimates of the variance at each target location. IDW promises no such thing, hence appears less dependent on assumptions. This does not imply that IDW is more robust or almost-as-good for simple tasks.

For the numbers that both methods do provide, it is possible to evaluate and compare the two using leave-one-out cross validation.

With just a reasonable a priori judgment for the range parameter, my guess is that Kriging will beat IDW in cross-validation for most problems, including notoriously assumption-breaking daily precipitation data. With some scripting, it is also possible to optimise Kriging's semivariogram parameters by minimising the cross-validation RMSE. This does not qualify for being BLUE or for trusting the variance estimates, though.

All this said, the extra burden of finding Kriging in one of QGIS' plugins, may serve as a warning that there is more to it than just plug-and play.

Sjur :-)

-----Original Message-----
Sent: 26. september 2015 19:43
To: Stefan Keller
Subject: Re: [Qgis-developer] Kriging interpolation functionality in QGIS?

Post by Stefan Keller
Hi Barry
Many thanks for your explanations and hints.
Do you think Inverse Distance Weighting (IDW) would do the job as
well, since Kriging has so many parameters to fiddle around and to

understand?
Depends on what "the job" is. To get an impression of the overall trend of a set
of samples - IDW is probably fine. But to get a
*principled* (ie based on a statistical model) set of estimates over a grid *with
honest estimates of uncertainty* so you can answer probabilistic questions (like
"what's the chance that the soil over there will contain 1ug gold/tonne?") you
need something like kriging.
Barry
_______________________________________________
Qgis-developer mailing list
http://lists.osgeo.org/mailman/listinfo/qgis-developer

[Victor Olaya]

Kriging is not that hard to find in QGIS, you just have to go to the
Processing search box and type "Kriging" to find all the the kriging
algorithms ;-)

From the point of view of user-friendliness, the Processing
implementation (wrapping the corresponding SAGA modules), might not be
as easy to use as some people would like, but for tools such as
kriging I am strongly against wizard-like UI's and similar elements.
ArcGIS's Statistical Analyst is great and has a wizard with a
fantastic "next" button that allows you to interpolate using all sort
of esoteric methods and will make you believe that you are creating
sound raster layers...when the truth is that, without knowledge, you
are creating rubish. I don't like to give users that wrong sensation.

People want a "Kriging for dummies" functionality, and that's is a bad
idea. Kriging is complex and it's not easy to understand the
underlying principles...but you have to understand them if you want to
use the tool. I think there's no other functionality as misused in GIS
software as this one...everyone want to use kriging because it's cool
and they have heard it's better...but without studying what it really
is.

(I get questions about this topic often, so here's my take on it and
on why the Processing implementation is like that, in case it might
help).

Regards

Post by Sjur Kolberg
I'd prefer Kriging to IDW also for simple jobs.
0: Know the assumptions and check your data, transform if necessary and possible.
1: Estimate your empirical semivariogram (describing how difference increases with distance).
2: Select a parametric semivariogram model, and adjust parameters to fit the point cloud from 1)
3: Use the parametric model to calculate a weight matrix between each target location (grid point) and each observation point.
Loosely speaking, the three first operations define the core of Kriging. The last operation answers the same question as IDW, both providing the coefficients (weights) in a linear combination of the input data.
With Gaussian data, spatial stationarity, well-spread data points and a properly selected/tuned semivariogram model, Kriging can be shown to be 'Best linear unbiased estimator' (BLUE), and provides mathematically correct estimates of the variance at each target location. IDW promises no such thing, hence appears less dependent on assumptions. This does not imply that IDW is more robust or almost-as-good for simple tasks.
For the numbers that both methods do provide, it is possible to evaluate and compare the two using leave-one-out cross validation.
With just a reasonable a priori judgment for the range parameter, my guess is that Kriging will beat IDW in cross-validation for most problems, including notoriously assumption-breaking daily precipitation data. With some scripting, it is also possible to optimise Kriging's semivariogram parameters by minimising the cross-validation RMSE. This does not qualify for being BLUE or for trusting the variance estimates, though.
All this said, the extra burden of finding Kriging in one of QGIS' plugins, may serve as a warning that there is more to it than just plug-and play.
Sjur :-)

-----Original Message-----
Sent: 26. september 2015 19:43
To: Stefan Keller
Subject: Re: [Qgis-developer] Kriging interpolation functionality in QGIS?

Post by Stefan Keller
Hi Barry
Many thanks for your explanations and hints.
Do you think Inverse Distance Weighting (IDW) would do the job as
well, since Kriging has so many parameters to fiddle around and to

understand?
Depends on what "the job" is. To get an impression of the overall trend of a set
of samples - IDW is probably fine. But to get a
*principled* (ie based on a statistical model) set of estimates over a grid *with
honest estimates of uncertainty* so you can answer probabilistic questions (like
"what's the chance that the soil over there will contain 1ug gold/tonne?") you
need something like kriging.
Barry
_______________________________________________
Qgis-developer mailing list
http://lists.osgeo.org/mailman/listinfo/qgis-developer

_______________________________________________
Qgis-developer mailing list
http://lists.osgeo.org/mailman/listinfo/qgis-developer

[René-Luc Dhont]

Hi,

You'll find 2 krigings algorithm implemented in R.

In the processing toolbox, you can get these scripts with R scripts >
Tools > Get R scripts from on-line scripts :
* Kriging
* Kriging with model selection

Regards,

Post by Victor Olaya
Kriging is not that hard to find in QGIS, you just have to go to the
Processing search box and type "Kriging" to find all the the kriging
algorithms ;-)
From the point of view of user-friendliness, the Processing
implementation (wrapping the corresponding SAGA modules), might not be
as easy to use as some people would like, but for tools such as
kriging I am strongly against wizard-like UI's and similar elements.
ArcGIS's Statistical Analyst is great and has a wizard with a
fantastic "next" button that allows you to interpolate using all sort
of esoteric methods and will make you believe that you are creating
sound raster layers...when the truth is that, without knowledge, you
are creating rubish. I don't like to give users that wrong sensation.
People want a "Kriging for dummies" functionality, and that's is a bad
idea. Kriging is complex and it's not easy to understand the
underlying principles...but you have to understand them if you want to
use the tool. I think there's no other functionality as misused in GIS
software as this one...everyone want to use kriging because it's cool
and they have heard it's better...but without studying what it really
is.
(I get questions about this topic often, so here's my take on it and
on why the Processing implementation is like that, in case it might
help).
Regards

Post by Sjur Kolberg
I'd prefer Kriging to IDW also for simple jobs.
0: Know the assumptions and check your data, transform if necessary and possible.
1: Estimate your empirical semivariogram (describing how difference increases with distance).
2: Select a parametric semivariogram model, and adjust parameters to fit the point cloud from 1)
3: Use the parametric model to calculate a weight matrix between each target location (grid point) and each observation point.
Loosely speaking, the three first operations define the core of Kriging. The last operation answers the same question as IDW, both providing the coefficients (weights) in a linear combination of the input data.
With Gaussian data, spatial stationarity, well-spread data points and a properly selected/tuned semivariogram model, Kriging can be shown to be 'Best linear unbiased estimator' (BLUE), and provides mathematically correct estimates of the variance at each target location. IDW promises no such thing, hence appears less dependent on assumptions. This does not imply that IDW is more robust or almost-as-good for simple tasks.
For the numbers that both methods do provide, it is possible to evaluate and compare the two using leave-one-out cross validation.
With just a reasonable a priori judgment for the range parameter, my guess is that Kriging will beat IDW in cross-validation for most problems, including notoriously assumption-breaking daily precipitation data. With some scripting, it is also possible to optimise Kriging's semivariogram parameters by minimising the cross-validation RMSE. This does not qualify for being BLUE or for trusting the variance estimates, though.
All this said, the extra burden of finding Kriging in one of QGIS' plugins, may serve as a warning that there is more to it than just plug-and play.
Sjur :-)

-----Original Message-----
Sent: 26. september 2015 19:43
To: Stefan Keller
Subject: Re: [Qgis-developer] Kriging interpolation functionality in QGIS?

Post by Stefan Keller
Hi Barry
Many thanks for your explanations and hints.
Do you think Inverse Distance Weighting (IDW) would do the job as
well, since Kriging has so many parameters to fiddle around and to

understand?
Depends on what "the job" is. To get an impression of the overall trend of a set
of samples - IDW is probably fine. But to get a
*principled* (ie based on a statistical model) set of estimates over a grid *with
honest estimates of uncertainty* so you can answer probabilistic questions (like
"what's the chance that the soil over there will contain 1ug gold/tonne?") you
need something like kriging.
Barry
_______________________________________________
Qgis-developer mailing list
http://lists.osgeo.org/mailman/listinfo/qgis-developer

_______________________________________________
Qgis-developer mailing list
http://lists.osgeo.org/mailman/listinfo/qgis-developer

_______________________________________________
Qgis-developer mailing list
http://lists.osgeo.org/mailman/listinfo/qgis-developer

[Barry Rowlingson]

Post by Victor Olaya
From the point of view of user-friendliness, the Processing
implementation (wrapping the corresponding SAGA modules), might not be
as easy to use as some people would like, but for tools such as
kriging I am strongly against wizard-like UI's and similar elements.
ArcGIS's Statistical Analyst is great and has a wizard with a
fantastic "next" button that allows you to interpolate using all sort
of esoteric methods and will make you believe that you are creating
sound raster layers...when the truth is that, without knowledge, you
are creating rubish. I don't like to give users that wrong sensation.

Agreed. Anything claiming to be a "wizard" for Kriging really needs to
be a psychic wizard that can read your thoughts and understand your
data. A better device would be an "Interrogator":

*ping* Your data seems to be discrete small integers, are you sure you
want to model it as a Gaussian? [*yes*/no]

*ping* So you want to model your data with Gaussian cofactor response
p-value thresholds from an underlying Bayesian surface? [*yes*/no]

*ping* That last question was nonsense. You're just clicking "yes"
until you get a surface model - any surface model - right? [*yes*/no]

*ping* okay, at least you are honest. I'm deleting the kriging module
now, go find a statistician.

Barry

[Werner Macho]

Well written and I fully support that.
Sometimes it seems that people want a wizard for everything (and a
computer that can read their mind).
I am against imitating everything that "commercial software" provides.
Sometimes it is better to treat users to switch on their brain.
For me that is still the point that separates Opensource Software from
proprietary one.
I am not sure if we should try to get more users at all cost. Let's do
things correct - even if that is not always popular (because it is not
in reach with one click).

just my 2c
Werner

On Mon, Sep 28, 2015 at 1:37 PM, Barry Rowlingson

Post by Barry Rowlingson

Post by Victor Olaya
From the point of view of user-friendliness, the Processing
implementation (wrapping the corresponding SAGA modules), might not be
as easy to use as some people would like, but for tools such as
kriging I am strongly against wizard-like UI's and similar elements.
ArcGIS's Statistical Analyst is great and has a wizard with a
fantastic "next" button that allows you to interpolate using all sort
of esoteric methods and will make you believe that you are creating
sound raster layers...when the truth is that, without knowledge, you
are creating rubish. I don't like to give users that wrong sensation.

Agreed. Anything claiming to be a "wizard" for Kriging really needs to
be a psychic wizard that can read your thoughts and understand your
*ping* Your data seems to be discrete small integers, are you sure you
want to model it as a Gaussian? [*yes*/no]
*ping* So you want to model your data with Gaussian cofactor response
p-value thresholds from an underlying Bayesian surface? [*yes*/no]
*ping* That last question was nonsense. You're just clicking "yes"
until you get a surface model - any surface model - right? [*yes*/no]
*ping* okay, at least you are honest. I'm deleting the kriging module
now, go find a statistician.
Barry
_______________________________________________
Qgis-developer mailing list
http://lists.osgeo.org/mailman/listinfo/qgis-developer

[Stefan Keller]

Werner, Barry

I just wrote to have Kriging in QGIS with less external *dependencies*
like GRASS, SAGA (or R).
That's what I meant by "easier to use" in my initial thread.

Then I challenged the benefit of Kriging versus IDW just because I
don't want to use it because it's cool.
Thanks to Barry and others for the explanations.

The use case you obviously have in mind, is to do data analytics as a
researcher or informed user - which requires the steps Sjur mentioned
(many thanks too).

On the other hand, to me it's still worth thinking of a helper dialog
(in whatever implementation and whatever you call it) which suggests
parameters calculated on cross validation.

... Unless you are saying that only statisticians can ever master Kriging :-O

:Stefan

Post by Werner Macho
Well written and I fully support that.
Sometimes it seems that people want a wizard for everything (and a
computer that can read their mind).
I am against imitating everything that "commercial software" provides.
Sometimes it is better to treat users to switch on their brain.
For me that is still the point that separates Opensource Software from
proprietary one.
I am not sure if we should try to get more users at all cost. Let's do
things correct - even if that is not always popular (because it is not
in reach with one click).
just my 2c
Werner
On Mon, Sep 28, 2015 at 1:37 PM, Barry Rowlingson

Post by Barry Rowlingson

Post by Victor Olaya
From the point of view of user-friendliness, the Processing
implementation (wrapping the corresponding SAGA modules), might not be
as easy to use as some people would like, but for tools such as
kriging I am strongly against wizard-like UI's and similar elements.
ArcGIS's Statistical Analyst is great and has a wizard with a
fantastic "next" button that allows you to interpolate using all sort
of esoteric methods and will make you believe that you are creating
sound raster layers...when the truth is that, without knowledge, you
are creating rubish. I don't like to give users that wrong sensation.

Agreed. Anything claiming to be a "wizard" for Kriging really needs to
be a psychic wizard that can read your thoughts and understand your
*ping* Your data seems to be discrete small integers, are you sure you
want to model it as a Gaussian? [*yes*/no]
*ping* So you want to model your data with Gaussian cofactor response
p-value thresholds from an underlying Bayesian surface? [*yes*/no]
*ping* That last question was nonsense. You're just clicking "yes"
until you get a surface model - any surface model - right? [*yes*/no]
*ping* okay, at least you are honest. I'm deleting the kriging module
now, go find a statistician.
Barry
_______________________________________________
Qgis-developer mailing list
http://lists.osgeo.org/mailman/listinfo/qgis-developer

_______________________________________________
Qgis-developer mailing list
http://lists.osgeo.org/mailman/listinfo/qgis-developer

[Victor Olaya]

Stefan

Leaving aside the statistical discussion, the Processing
implementation doesn't use really aný "external" element. It uses
SAGA, that's true, but if you use a QGIS package (in Win or Mac), SAGA
is installed and configured automatically, and it is used as any of
the dozens of libraries that QGIS uses underneath ...so the user wont
notice that is using an external software, just like you don't notice
that you use GDAL for opening layers.

Using R is a differenet thing, that requires manual configuration...

Cheers

Post by Stefan Keller
Werner, Barry
I just wrote to have Kriging in QGIS with less external *dependencies*
like GRASS, SAGA (or R).
That's what I meant by "easier to use" in my initial thread.
Then I challenged the benefit of Kriging versus IDW just because I
don't want to use it because it's cool.
Thanks to Barry and others for the explanations.
The use case you obviously have in mind, is to do data analytics as a
researcher or informed user - which requires the steps Sjur mentioned
(many thanks too).
On the other hand, to me it's still worth thinking of a helper dialog
(in whatever implementation and whatever you call it) which suggests
parameters calculated on cross validation.
... Unless you are saying that only statisticians can ever master Kriging :-O
:Stefan

Post by Werner Macho
Well written and I fully support that.
Sometimes it seems that people want a wizard for everything (and a
computer that can read their mind).
I am against imitating everything that "commercial software" provides.
Sometimes it is better to treat users to switch on their brain.
For me that is still the point that separates Opensource Software from
proprietary one.
I am not sure if we should try to get more users at all cost. Let's do
things correct - even if that is not always popular (because it is not
in reach with one click).
just my 2c
Werner
On Mon, Sep 28, 2015 at 1:37 PM, Barry Rowlingson

Post by Barry Rowlingson

Post by Victor Olaya
From the point of view of user-friendliness, the Processing
implementation (wrapping the corresponding SAGA modules), might not be
as easy to use as some people would like, but for tools such as
kriging I am strongly against wizard-like UI's and similar elements.
ArcGIS's Statistical Analyst is great and has a wizard with a
fantastic "next" button that allows you to interpolate using all sort
of esoteric methods and will make you believe that you are creating
sound raster layers...when the truth is that, without knowledge, you
are creating rubish. I don't like to give users that wrong sensation.

Agreed. Anything claiming to be a "wizard" for Kriging really needs to
be a psychic wizard that can read your thoughts and understand your
*ping* Your data seems to be discrete small integers, are you sure you
want to model it as a Gaussian? [*yes*/no]
*ping* So you want to model your data with Gaussian cofactor response
p-value thresholds from an underlying Bayesian surface? [*yes*/no]
*ping* That last question was nonsense. You're just clicking "yes"
until you get a surface model - any surface model - right? [*yes*/no]
*ping* okay, at least you are honest. I'm deleting the kriging module
now, go find a statistician.
Barry
_______________________________________________
Qgis-developer mailing list
http://lists.osgeo.org/mailman/listinfo/qgis-developer

_______________________________________________
Qgis-developer mailing list
http://lists.osgeo.org/mailman/listinfo/qgis-developer

_______________________________________________
Qgis-developer mailing list
http://lists.osgeo.org/mailman/listinfo/qgis-developer

[Anita Graser]

Hi Victor,

Post by Victor Olaya
but if you use a QGIS package (in Win or Mac), SAGA
is installed and configured automatically, and it is used as any of
the dozens of libraries that QGIS uses underneath ...

​What's the correct SAGA path for a OSGeo4W64 install (relative to the
OSGeo4W64 folder)? On my system, the SAGA path is not configured
automatically in the Processing settings and log says "Problem with SAGA
installation: SAGA was not found or is not correctly installed".

Thanks a lot and best wishes,
Anita​

[Anita Graser]

Post by Anita Graser
Hi Victor,

Post by Victor Olaya
but if you use a QGIS package (in Win or Mac), SAGA
is installed and configured automatically, and it is used as any of
the dozens of libraries that QGIS uses underneath ...

​What's the correct SAGA path for a OSGeo4W64 install (relative to the
OSGeo4W64 folder)? On my system, the SAGA path is not configured
automatically in the Processing settings and log says "Problem with SAGA
installation: SAGA was not found or is not correctly installed".

​Ok, sorry for the noise. I had to install SAGA separately in the OSGeo4W
installer because it seems like it's not automatically picked as a
dependency.

Should I open a (OSGeo4W?) ticket or is this behavior on purpose?

Best wishes,
Anita​

[Werner Macho]

Hi Stefan,

My post was in no way meant offending. I replied to Victor and Barry
just because of the nice joke about "wizards".
My experience at university is just that a lot of students use
"wizards" without knowing what they are doing and Barrys post
described exactly what I saw at university.

I hope you will find (or already found) what you are looking for.
With Kriging unfortunately I cannot help you because I just did not
understand it ;)

regards,
Werner

Post by Stefan Keller
Werner, Barry
I just wrote to have Kriging in QGIS with less external
*dependencies* like GRASS, SAGA (or R). That's what I meant by
"easier to use" in my initial thread.
Then I challenged the benefit of Kriging versus IDW just because I
don't want to use it because it's cool. Thanks to Barry and others
for the explanations.
The use case you obviously have in mind, is to do data analytics as
a researcher or informed user - which requires the steps Sjur
mentioned (many thanks too).
On the other hand, to me it's still worth thinking of a helper
dialog (in whatever implementation and whatever you call it) which
suggests parameters calculated on cross validation.
... Unless you are saying that only statisticians can ever master Kriging :-O
:Stefan

Well written and I fully support that. Sometimes it seems that
people want a wizard for everything (and a computer that can read
their mind). I am against imitating everything that "commercial
software" provides. Sometimes it is better to treat users to
switch on their brain. For me that is still the point that
separates Opensource Software from proprietary one. I am not sure
if we should try to get more users at all cost. Let's do things
correct - even if that is not always popular (because it is not
in reach with one click).
just my 2c Werner
On Mon, Sep 28, 2015 at 1:37 PM, Barry Rowlingson

On Mon, Sep 28, 2015 at 10:35 AM, Victor Olaya

Post by Victor Olaya
From the point of view of user-friendliness, the Processing
implementation (wrapping the corresponding SAGA modules),
might not be as easy to use as some people would like, but
for tools such as kriging I am strongly against wizard-like
UI's and similar elements. ArcGIS's Statistical Analyst is
great and has a wizard with a fantastic "next" button that
allows you to interpolate using all sort of esoteric methods
and will make you believe that you are creating sound raster
layers...when the truth is that, without knowledge, you are
creating rubish. I don't like to give users that wrong
sensation.

Agreed. Anything claiming to be a "wizard" for Kriging really
needs to be a psychic wizard that can read your thoughts and
understand your data. A better device would be an
*ping* Your data seems to be discrete small integers, are you
sure you want to model it as a Gaussian? [*yes*/no]
*ping* So you want to model your data with Gaussian cofactor
response p-value thresholds from an underlying Bayesian
surface? [*yes*/no]
*ping* That last question was nonsense. You're just clicking
"yes" until you get a surface model - any surface model -
right? [*yes*/no]
*ping* okay, at least you are honest. I'm deleting the kriging
module now, go find a statistician.
Barry _______________________________________________
http://lists.osgeo.org/mailman/listinfo/qgis-developer

_______________________________________________ Qgis-developer
http://lists.osgeo.org/mailman/listinfo/qgis-developer

[Stefan Keller]

Werner: I'm with you.
Victor: Thanks for the pointer to SAGA tools in Processing.

The background of my questions is, that I'm having a research project
in mind where we need to implement Kriging for predefined agricultural
soil samples.
Here the farmer does not need to know what's under the hood.
Now I'm thinking to give back something to the QGIS community like
parameter suggestions calculated from cross validation.
And I'm reluctant to implement this using something other than a QGIS
extension (nothing against SAGA).

Any one who has experiences on how estimate any Kriging parameters?
E.g. these SAGA tool parameters look especially "frightening" :-)
Power function A/B (in Universal K.) and Variogram Model (in Ordinary K.)

:Stefan

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Hi Stefan,
My post was in no way meant offending. I replied to Victor and Barry
just because of the nice joke about "wizards".
My experience at university is just that a lot of students use
"wizards" without knowing what they are doing and Barrys post
described exactly what I saw at university.
I hope you will find (or already found) what you are looking for.
With Kriging unfortunately I cannot help you because I just did not
understand it ;)
regards,
Werner

Post by Stefan Keller
Werner, Barry
I just wrote to have Kriging in QGIS with less external
*dependencies* like GRASS, SAGA (or R). That's what I meant by
"easier to use" in my initial thread.
Then I challenged the benefit of Kriging versus IDW just because I
don't want to use it because it's cool. Thanks to Barry and others
for the explanations.
The use case you obviously have in mind, is to do data analytics as
a researcher or informed user - which requires the steps Sjur
mentioned (many thanks too).
On the other hand, to me it's still worth thinking of a helper
dialog (in whatever implementation and whatever you call it) which
suggests parameters calculated on cross validation.
... Unless you are saying that only statisticians can ever master Kriging :-O
:Stefan

Well written and I fully support that. Sometimes it seems that
people want a wizard for everything (and a computer that can read
their mind). I am against imitating everything that "commercial
software" provides. Sometimes it is better to treat users to
switch on their brain. For me that is still the point that
separates Opensource Software from proprietary one. I am not sure
if we should try to get more users at all cost. Let's do things
correct - even if that is not always popular (because it is not
in reach with one click).
just my 2c Werner
On Mon, Sep 28, 2015 at 1:37 PM, Barry Rowlingson

On Mon, Sep 28, 2015 at 10:35 AM, Victor Olaya

Post by Victor Olaya
From the point of view of user-friendliness, the Processing
implementation (wrapping the corresponding SAGA modules),
might not be as easy to use as some people would like, but
for tools such as kriging I am strongly against wizard-like
UI's and similar elements. ArcGIS's Statistical Analyst is
great and has a wizard with a fantastic "next" button that
allows you to interpolate using all sort of esoteric methods
and will make you believe that you are creating sound raster
layers...when the truth is that, without knowledge, you are
creating rubish. I don't like to give users that wrong
sensation.

Agreed. Anything claiming to be a "wizard" for Kriging really
needs to be a psychic wizard that can read your thoughts and
understand your data. A better device would be an
*ping* Your data seems to be discrete small integers, are you
sure you want to model it as a Gaussian? [*yes*/no]
*ping* So you want to model your data with Gaussian cofactor
response p-value thresholds from an underlying Bayesian
surface? [*yes*/no]
*ping* That last question was nonsense. You're just clicking
"yes" until you get a surface model - any surface model -
right? [*yes*/no]
*ping* okay, at least you are honest. I'm deleting the kriging
module now, go find a statistician.
Barry _______________________________________________
http://lists.osgeo.org/mailman/listinfo/qgis-developer

_______________________________________________ Qgis-developer
http://lists.osgeo.org/mailman/listinfo/qgis-developer

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2
iEYEARECAAYFAlYJlsMACgkQDAH1YiCxBgkjCgCgg+dpbAzdssktln4MgSQTCqoP
XX4An0HoPweuDt0wyv7S6jrOMAlkTTAp
=iyga
-----END PGP SIGNATURE-----