paladin74 / geoda Goto Github PK

from  Mar 7, 2011

Would be great to offer users the option to view and edit the weights.gal/.gwt 
files within OG instead of having to go outside to Notepad etc.

May be one option would be to display the gal/gwt files as a table with rows 
and columns so it's possible to label both and to link the core and neighboring 
observations to those in the map when they are selected in the table (or vice 
versa, in the map).

from Aug 24, 2012

CSV import currently fails on blank values such as:

1, 2, 3
4,, 6

from Aug 18, 2012

It would be useful to have a graph that allows users to compare the significant 
z-values associated with different Moran scatterplots.

For the time case, X would be time and Y the significant z-values. In other 
cases, Y would still be sig. z-values but X could be different Ws (to allow 
users to compare how sensitive Moran's I is for different W specifications).

from Aug 18, 2012

Currently, the global Moran's I statistic for spatial autocorrelation we have 
in GeoDa is linear. Luc has been interested in adding nonlinear tests and 
displays (e.g. copula, lowess/locally weighted scatterplot smoothing).

from Jan 18, 2011

The Progress Dialog when opening a new shp file is slow to update on Windows 
and especially on Linux.  I have tried forcing refreshes and updates to the 
window, but it makes no difference.  The embedded progress bar in the 
Regression Dialog seems to work well on all platforms, but the text on Linux 
does not update well.

from Oct 28, 2009

Add ability to add columns for area and perimeter calculations.  The area
calculation in particular would be useful for calculating population density.

from Oct 12, 2011

On Windows in Linux, the zoom out tool shows a + rather than a - as would be 
expected.  In OSX the zoom in/out tools are just a simple magnifying glass, so 
there's no issue.

from Dec 28, 2011

We're getting a lot of people with questions based on older versions of OG that 
are resolved in the latest version but people didn't update.

Mark suggested that it would be possible to add an automatic update checker 
into GeoDa to let people know that a newer version is available.

Would be great to have this since we update frequently.

Please update the Regression report window with HTML tables and a
variable-width font.

(from Sep 9, 2009)

from Oct 29, 2009

We could add a mode where certain weights matrices are computed in
real-time.  For example, we could have a queen or rook mode contiguity
weights mode were the weights matrix would be automatically adjusted as
certain observations are added or removed.  This would allow for the
correct recomputation of Moran's I when certain observations are removed.

from  Sep 19, 2011

Some users are confused by the current way that multiple Regression Reports are 
appended together---sometimes users think that later models were never run.  To 
prevent this, I suggest always reporting the latest model at the top, or else 
automatically scrolling the scroll window to the bottom.

from Aug 18, 2012

It would be useful to have a graph that allows users to compare the betas and 
standard errors associated with scatterplots over time to eyeball if there are 
significant differences btw the betas over time.

This could e.g. be similar to your display of boxplots for a variable over time 
except that only the beta value (like the green dot) and the standard error 
whiskers would be shown.

When a non-valid table file is chosen for "Open Table Only" such as a .shp file 
instead of a .csv or .dbf file, the program crashes.  In general, we should 
handle error situations like this better.

I believe Legacy Geoda used to have a way to use an irregular polygon as the 
selection shape, by clicking a number of times on the map. It may be useful to 
reinstate this as one of the selection options, since the rectangle or circle 
may be too limiting.

from Oct 13, 2011

Add a feature that does a fast auto-column width adjustment.  The built-in 
Table auto-column-width function is too slow for large Tables.  We can make a 
feature that samples the first and last Table items as well as some number of 
randomly chosen rows to guess a good column width size.  This function should 
be called automatically when a new table is loaded.

from Sep 9, 2009

In many dialog windows, the options are OK and Close, but normally there
should be a button to Cancel the current operation. 

from  Sep 29, 2011

This comes from Richard on the OpenSpace list.  Here's the original suggestion:

When making a percentile map it would be useful to be able to see what the 
actual values are for the data in the legend, so if I see 10%-50% (528) (the 
percentile and the number of polygons in that percentile range, being able to 
see the data value boundaries, say 10%-50% (528) 33.45-67.01  would be very 
helpful.


I'd also suggest that we add the ability to click on legend categories to 
select all observations in that particular category.  A user could then use the 
Table to see what the category boundaries are.

In GeoDa 1.4.3, when right clicking "Display Statistics" in Scatterplot, the 
plot does not update until the window is clicked again.  Problem seen on OSX, 
but not tested on other platforms.

from Oct 13, 2011

Because of the limitation of the DBF fixed-precision decimal format, it very 
easy to generate values in field calculations that will be truncated when 
saving the DBF back to disk.  We should have a way to detect this and to warn 
the user.  We could also give the option to slightly modify the field 
properties for the problematic field to prevent the truncation when possible.

Oct 28, 2009

It would seem to be useful to allow for infinity or undefined in numeric
data.  This could be particularly useful for handling divide-by-zero or
network distance calculations.  For the network distance calculation, we
could possibly avoid infinity by disallowing certain calculations on
disconnected graphs.

Update Mar 5, 2013: As of 1.2 we do handle undefined.  Due to limitation in 
DBF, it is not possible to distinguish between infinity and undefined.

(from Sep 9, 2009)
additional spatial lags: not just average but also sum and conditional sum,
in case people want totals or want to compute rates or lagged rates (as
requested by jkoschinsky).

This could be part of the table refactoring. May be it makes sense to do one 
ticket: Refactor Table and then put all the small table enhancements in as 
subsets?

the default circle size in the bubble chart seems to be too large.
the bubbles are too overwhelming and there is too much overlap.
need consider a scaling algorithm.

from Oct 13, 2011

The new Table is now aware of undefined values.  These can easily be produced 
when a field calculation does a divide-by-zero or produced an imaginary number. 
 We should display a warning dialog to the user when this happens.  The warning 
dialog should also include a checkbox option to suppress further warnings.

Instead of just displaying warnings for undefined values, adjust every view to 
account for undefined values (e.g. exclude them from calculation and may be 
display in separate legend category (similar to what we're doing with islands 
in LISAs) but still allow for selection).

from Sep 18, 2009

Point data is highlighted by changing the outline color.  The issue is that
for certain coloring schemes, some of the points essentially disappear
against the black background.  Perhaps would make sense to have point
outline in black, and change the inside color.

from Aug 17, 2012
Suggested solution: fill in circle color (right-click instead of default for 
performance?) and allow for outlines to be turned on.

Similar to the logic used for the scatter plot, we could implement a test on 
difference in means (a simple z-stat) between selected and unselected 
observations in PCP, for each of the variables included.

A related idea is to visualize the summary statistics that are now given, i.e., 
show a line that corresponds to the mean values for each variable.

(from Sep 9, 2009)
e.g., New (cmd-N), Close window (cmd-W), open (cmd-o), maybe show/hide
results (cmd-r)

Update: We now of have CMD-W for close window and others will be added
soon. Could possibly add CMD-C for copy to clipboard?

Table Merge dialog needs to be updated to support OGR sources.  Additionally, 
we will add the ability to do left/right and full joins as needed when row 
mismatches are present.  We will present the user with helpful dialogs 
explaining the options as well as useful metrics for making their decision.  
For example, if the row mismatch is more than 5%, then present a warning.

(from Sep 2, 2009)
In Linux, default directories aren't remembered.  Remember them or pick
more sensible defaults based on the currently opened shape file.

currently, the default for the second variable in the scatter plot is whatever 
was selected as a second variable before. it would be better to set the second 
variable automatically to match the first variable, unless the
user changes it.

so, the default for the second variable (color) would be to always match the 
first variable, and to follow changes in the first variable. it would require 
an explicit specification from the user to set the second variable to be 
different.

as it is, the default for the second variable right now is almost always 
different from the first, which often leads to unintended results

from Aug 6, 2012

beta 40, Mac

Global Moran's I randomization window: doesn't scale properly for lower 
resolutions.

A GeoDa user reported this issue to us and I can reproduce it on GeoDa 1.2.0 
and 1.3.27.  Steps to reproduce the problem:

1. Open a Rates Thematic Map (I did Excess Risk)
2. Choose Save Rates
3. Choose Create New Variable
4. Choose option to add variable to "end of table"

Program will crash after selecting Apply in "Save Field To Table" window.

This only happens on Windows, but will likely cause future problems on Mac.  
Haven't tested on Linux.

In 1.4.3, when choosing a Quantile map, and specifying 6 categories (for 
example), the number of categories in any other variable # category should 
default to 6.  Currently, it always defaults back to 4.

Add ability to convert data type.

When selecting a different classification theme in PCP (right click > 
Classification Themes), the new
selection immediately shows up in the left hand legend panel, but it requires 
an extra click on the
right panel (with the PCP) to implement it. So, without the extra click, there 
is a disconnect between
the legend (new classification) and the graph (old classification). The change 
should be instantaneous
in both.

from Sep 16, 2009

for example, when creating a grid shape file, there is no success message.

Aug 18, 2012

Many OG users work with Census data with FIPS codes in string format as the 
unique ID because these codes often have leading zeros. Let's look at what it 
would take to allow users to specify unique strings as IDs for the weights 
dialog.

from Sep 14, 2011

Data: Baltimore
Win ..15.7

Last digit differs in spatial lag model (they match in spatial error model) - 
not sure we care but wanted to log it in case we do:

Regression.OLS = Legacy
baltRegression.txt = OG ..15.7

OG - Spatial Lag
CONSTANT      5.950557       4.830249       1.231936    0.2179732
       LOTSZ    0.09989266     0.01685041       5.928204    0.0000000

Legacy
    CONSTANT      5.950559       4.830249       1.231936    0.2179730
       LOTSZ    0.09989267     0.01685041       5.928204    0.0000000

Another note: In OG the Regression Report viewer stops displaying text after a 
while, so if output is long it won't display the whole thing (but all output is 
saved if people want to view it outside).

from Sep 15, 2011
These differences are only in the 7th decimal place, and even then, the 
difference is at most by two thenths in that place.  Some of the math libraries 
in GeoDa are different than those used in Legacy, so I think that differences 
of this magnitude should be expected.  Moving to future.

from Oct 29, 2009

Add the option to assign centroid and mean-center x/y values to a field. 
Could also consider adding simple neighbor measures such as number of rook
and queen neighbors.

from Jun 7, 2012

When opening a Space-Time shapefile, the user can choose the shared TIME id 
field.  Normally, the default value is correct.  However, if some other 
incorrect field is chosen, the program crashes rather than detecting the 
problem.

from Sep 16, 2009

... rather than just cutting them off.

Update compile process GeoDa 1.5 for Linux.

from Sep 14, 2009

Add option for inverse distance in weights creation dialog.

This option was removed because it was not working correctly.

Part of larger weights refactoring.

an idea for the 3D scatterplot after seeing Ross' program where he uses 
shortest-distance cluster algorithms (like k-means and others) to identify 
similarity of multivariate coordinates. We could add something like this as a 
checkbox option to the 3 variables specified in our 3D scatterplot and then 
color-code any resulting cluster in the 3D scatter space and the map. Could 
also add a third X to the weights since we already allow users to specify 
non-geographic XYs in the weights dialog.

from Dec 7, 2010

Suggestion form one of our users: The ability to select by SQL query of the 
.dbf could be added to the table.

from  Aug 7, 2012

beta 40

Zooming doesn't seem to be working quite fully yet: E.g. on Mac OSX when you 
try to zoom into the NCOVR map, it works three times and then no longer. 
Workshop participants encountered this in Windows, too.

from Aug 7, 2012

This limitation is currently intentional to limit the ammount of memory issued 
in the window buffering system.  There are designs that can get around this, 
but this will be a medium to larger effort enhancement.

the current defaults for the bubble chart third and fourth variable are
potentially misleading. they are set to the first variable in the list.

i suggest that the default setting for this is the same as for the
scatter plot, with the third and fourth variable only activated if
a variable is actively selected (if that's possible). if that's not
possible, then the third and fourth variable should default to
either the x or y variable. neither of these is ideal, but it would
be better than the current setting.

also, as suggested for the cartogram, the third and fourth variable
(size and color) should always default to be the same, and the 
fourth variable would need to be set explicitly.

[deleted issue]

from  Aug 18, 2012

One of the problems with the current PCP display for larger datasets is that 
larger values are drawn on top of smaller ones, giving a distorted visual 
impression. One solution Mark suggested to try instead (or as an additional 
display option) is to have the color scheme of the lines be based on a random 
sample to give a more representative visual impression of the values in the 
dataset.

from  Aug 18, 2012

Light point outline colors are currently hard to see on the default white 
backgrounds. One solution is for users to change the background color. 

We discussed alternatives, e.g. coloring the circle body, although this causes 
performance hits. Logging this issue as a reminder.

from  Oct 14, 2009

Time to add some unit testing to OpenGeoDa.  I propose to use the Boost
unit testing framework (UTF).  See
http://www.boost.org/doc/libs/1_40_0/libs/test/doc/html/index.html