The problem with your code is that $i are indexes into @arr, but you use it to delete rows from @arr1.

That works for the first row to be deleted, but afterwards the indexes differ. Inserting $i--; directly after the call to splice might fix it (haven't tried it though).

Update: this script seems to work:

use strict;
use warnings;

my @arr=(['a','b','c','d'],['e','f','g'],['h','i','j'],['e','f','g'],[
+'h','i','j'],['e','f','g'],['h','i','j'],['a','b','c','d'],['a','b','
+c','d']);
my @copy = @arr;

my $i=-1;

for(@arr){
    $i++;
    if(@$_[0] =~ /e/){
        splice(@copy,$i,1);
        $i--;
    }
}
print join (' ', @$_), "\n" for @copy;
[download]

You could also try a non-destructive approach:

my @arr1 = grep { @$_ !~ /e/ } @arr;
[download]

That is, construct a new array which only those rows you actually want.