EViews.com

Posted: **Sun Jul 31, 2011 9:53 am**

hey there,

i'm trying to program non-parametric tests whether a certain set of observations is significantly different from zero.

in order to program the simple sign test and wilcoxon and corrado rank sign tests, i need to solve the following issue:

suppose i have the following series grouped (in reality they're many more):

rs1 rs2 rs3 rs4
+1 -2 +2 +3
-2 -3 -1 -1
+2 +4 -1 -2
...

can anybody think of a way to count the number of positive items in a row?

what i need to have is a series containing these numbers (concerning all the series in the group) which would look like this for my example:
pos
3
0
2
...

thankful for any advice!

Posted: **Mon Aug 01, 2011 9:17 am**

The only way I can think is to create a new group with a @recoded version of each series and they count the number of observations in each row. You can do this in a small program without too much effort. Suppose group GRP1 has your series and that there is no group GRP2 in the workfile...

Code: Select all

group grp2
for !i=1 to group01.@count
%name = group01.@seriesname(!i)
grp2.add @recode({%name}>0, {%name}, na)
next
series numpos = @robs(grp2)
delete grp2

Here, we create a copy of the original group with new series that are based on the old, but with NAs whenever the series value is non-positive. Then we count the number of valid observations.

Posted: **Mon Aug 01, 2011 3:37 pm**

hey glenn! marvelous, that worked very well...so i'm done with programming all statistical tests i need thanks to you guys' help!
thanks a lot!!!!! :)

EViews.com

count positive group items

count positive group items

Re: count positive group items

Re: count positive group items