Different outputs for freq and freq +d3 commands

[Anne-Christin Tannhäuser]

I have a remark to make concerning the freq and freq +d3 (giving out
the stats, only) commands: In my analysis seems to be a constant flaw
concerning the freq +d3 output, i.e. all type counts are one item
higher than indicated by the freq analysis. This becomes especially
obvious, when using word lists to check for the occurence of certain
items and some data files contain none or only a few of those
particular items on the list. Then the TypeTokenRatio of the freq +d3
output will eventually become higher than 1.
I double and tripple checked the outputs and always ended up with the
same mistake. Has anybody encountered the same problem?

Kind Regards,
Anne-Christin

PS: Thank you Brian McWhinney for commenting on adjusting the header
lines of older CHAT files to the latest format.

[Leonid Spektor]

Anne-Christin,

Could you please give us command lines you use and email sample files
that demonstrate this problem. Also, tell us which Computer (PC, Mac PPC,
Mac Intel) and Operating system (Window 2000, XP, Vista, OSX 10.x) you are
using.

Thanks,
Leonid.

[Anne-Christin Tannhäuser]

Hello Leonid,

here the information you asked for:

command for calculation in CLAN output window:
freq +s@nouns. +t*SBJ

command for output in file:
freq +s@nouns. +t*SBJ @ +d3
and then
statfreq stat.out.cex +f +d

I use a PC and Windows XP.
Our data does not have a complete %mor line and therefore we used a
list of nouns to count their occurence in each file.
I email you the language data files, the noun list, plus the 2
different outputs I got.

Thank you so much in advance,
Anne-Christin

> > lines of older CHAT files to the latest format.- Hide quoted text -
>
> - Show quoted text -

[Leonid Spektor]

Anne-Christin,

Thanks for the sample file and exact instructions of how to replicate
the problem. I was able to find the bug. It occurred only when +d2 or +d3
option was used, i.e. statfreq program was involved. I have fixed this and
new CLAN is on childes web site.

Leonid.