Support for multiple 'percentile' arguments in r.univar

[epatton]

I'd like to submit a wish for having r.univar accept multiple percentile arguments as a part of its extended statistics, eg:

r.univar -e map=rast percentile=16,33,50,66,83

Thanks,

~ Eric.

[hamish]

yes, it would be nice to have that.

note 1st quartile, median, 3rd quartile give 25,50,75% already.

Hamish

[martinl]

'Percentile' parameter can be now multiple (also for r3.univar).

​http://trac.osgeo.org/grass/changeset/29848

E.g.

r.univar elevation.dem -e per=70,90
...
1st quartile: 1200
median (odd number of cells): 1316
3rd quartile: 1488
70th percentile: 1447
90th percentile: 1621

Martin

[hamish]

bug in the last change? r3.univar, rows*cols*depths, not rows*cols:

@@ -111,6 +113,12 @@
 
     map_type = G3d_tileTypeMap(map);
-    stats = create_univar_stat_struct(map_type, cols * rows * depths);
-    sscanf(param.percentile->answer, "%i", &stats->perc);
+
+    i = 0;
+    while(param.percentile->answers[i])
+	i++;
+    stats = create_univar_stat_struct(map_type, cols * rows, i);
+    for(i = 0; i < stats->n_perc; i++) {
+	sscanf(param.percentile->answers[i], "%i", &stats->perc[i]);
+    }

?? Hamish

[hamish]

and I'm not sure about this- if the user asks for 25,50,75 then give it to them, even if it is redundant info. It could be used as a verification test of sorts for those....?

stats.c

222               if (stats->perc[i] == 25 || 
223                    stats->perc[i] == 50 || 
224                    stats->perc[i] == 75) { 
225                    /* skip 1st quartile, median, 3rd quartile */ 
226                    continue; 
227               }

Hamish

[hamish]

if the user asks for 25,50,75 then give it to them

(my earlier post was meaning you might be able to re-use those values to save calculation time. the danger of adding complexity probably negates that savings though)

Also, any ideas about how to do FP percentiles? see ​http://article.gmane.org/gmane.comp.gis.grass.user/20744

In particular I am not sure how to solve problem 2.

thanks, Hamish

[martinl]

Replying to hamish:

bug in the last change? r3.univar, rows*cols*depths, not rows*cols:

@@ -111,6 +113,12 @@
 
     map_type = G3d_tileTypeMap(map);
-    stats = create_univar_stat_struct(map_type, cols * rows * depths);
-    sscanf(param.percentile->answer, "%i", &stats->perc);
+
+    i = 0;
+    while(param.percentile->answers[i])
+	i++;
+    stats = create_univar_stat_struct(map_type, cols * rows, i);
+    for(i = 0; i < stats->n_perc; i++) {
+	sscanf(param.percentile->answers[i], "%i", &stats->perc[i]);
+    }

sure, sorry, I fixed in trunk now.

Martin

[hamish]

ok to close this bug?

see changesets r29848, r29891, r29904

Hamish