Impatient user's guide to running analyses
[Return to the Contents
List]
Contents List
Getting
overview results NOW
Login to GALAXY on A6:: (password ZZZZZZ)
At the Huh? prompt, type (ASP, CH, EG,
EL, JB, JG, JH, JO, KH, KW, MC, MF, PB, PM, QT, RG, RKH, RUA, SD, SJ, SMR,
TE, VE, YW, ZMC etc as appropriate) <return> - this selects your
private directory [GALAXY.ASP] (or [GALAXY.KW] etc) where life can
be conducted undisturbed. Each private directory has an appropriate logical
name, e.g. ASP:
The 'default' overview is BC (type BC <return>
to switch back if you have been using others, e.g. CRC, HD etc).
To use the 'third frozen' data type BC$ICE3, to use the 'BC 1995' data
type BC$KK2, or to use the 'BC 1990' data type BC$KK.
For forest plots need to do 1,2,3 (patient
data > intermediate file > graphics); for life table curves, 1,2,4 (same
story). Of course, files might already exist from previous graunches. Beware
that things go out of date on timescales of hours, as new data and updates
are being put in! General remarks: dots are 'placeholders' in parameter
lists and usually mean 'nothing'. Filenames as parameters don't need suffices
(i.e. don't bother to put filename.DAT).
1. Check the steering file for the relevant
treatment comparison; all these are in BC$HB (or BC$HB:*.ICE3, or BC$KK2$HB
etc)
and are fairly self-explanatory. PLEASE DON'T alter the existing ones!!!
If change(s) needed, please copy into a 'clone' and use that. For instance,
if you wish to alter BC$HB:TAM12.DAT please copy it to e.g. BC$HB:TAM12EH.DAT
and use that.
2. To create the intermediate file containing
the annual events, at risk, O-E and Vs:
Huh? BLACK steeringfileendpointstratification
(...additional parameters)
N.B: for a 'large' question such as
TAM12, this can be time-consuming. Better to run as a subprocess (prefix
by RS, e.g. RS BLACK TAM12DD50) or in batch (prefix by RR, e.g.
RR BLACK TAM12DD50).
3a. To produce a forest plot metafile:
Huh? GAZEBO steeringfileendpointstratification
(...additional parameters)
- where steeringfileendpointstratification
is e.g. TAM12DD50
... this example would create a graphics metafile
TAM12DD50V.DAT
3b. To print the forest plot metafile (on
GRAFLP):
4a. To produce a life table plot metafile:
Huh? ABSEIL steeringfileendpointstratification
(...additional parameters)
- where steeringfileendpointstratification
is e.g. TAM12DD50
... this example would create a graphics metafile
TAM12DD50V.DAT
4b. To print the life table plot metafile
(on GRAFLP):
5. The summary forest plots (one line per
stratum) are a two-stage process:
Huh? GAZEBO TAM12DDA . . . . . . 1 (...other
parameters could be used in lieu of the dots)
... this example would create an intermediate
file TAM12DDAX.DAT (and a graphics metafile TAM12DDAV.DAT)
Huh? GAZEBO TAM12DDAX -1
... this would create a graphics metafile
TAM12DDAXV.DAT
Huh? LASER TAM12DDAXV
... this would print the summary forest
plot (LASER TAM12DDAV to print the one-page-per-stratum version as well)
6. The intermediate file for endpoint DDPOSTREC
is formed by subtraction between two intermediate files, previously created
by BLACKBOX, for the endpoints DD and NONBC. To do this:
Huh? POSTER steeringfile stratification
- where steeringfile is e.g. TAM12
stratification is e.g. 50
... this example would create an intermediate
file BC$DK:TAM12DDPOSTREC50.DAT, having previously run BLACKBOX for TAM12DD50
and TAM12NONBC50.
94. HELP - most GALAXY things are in HELP.
Also, most programs will give you useful information if you give a question
mark to them, e.g. BLACK ? There is a document in JG$DK:OVERVIEW.DAT
giving fuller details of file locations, software etc.
99. Various explanatory documents attached
- lists of file names, purposes and locations; strata and endpoints; definitions
of the data processing operations; other programs etc etc. Enjoy!
Changing programs
101. Most GALAXY commands call .EXE programs
from the prompt; the sources JG$AK:ABSEIL.FOR, JG$AK:BLACKBOX.FOR and JG$AK:GAZEBO.FOR
can easily be cloned and hacked (PLEASE DON'T alter the originals!) For
example, PB might create a new PB:XABSEIL.FOR, which can be compiled and
linked by typing
Huh? OO XABSEIL
... and then run by typing
Huh? ZAP XABSEIL ...parameters
What's
in GALAXY? - the files
GALAXY is a 'multi-user environment' which
contains (amongst other things) overview data and analysis systems for:
-
Breast Cancer (EBCTCG)
-
Colorectal Cancer
-
Hodgkin's Disease
-
Prostate Cancer
For each overview, the files held in GALAXY
include the following:
-
Individual patient data (as supplied by trialists,
in a very wide range of formats)
-
Individual patient data in 'standard format',
constructed from the stuff provided by the trialists
-
Update files of the individual patient data
(where additions and/or corrections have been made, or successive versions
have been provided by the trialists)
-
A 'master list' of the entire overview, containing
every trial and trial substratum of every participating group/centre: this
includes the precise details of every randomised treatment arm, eligibility
criteria, range of entry dates, total accrual, names of trialists, whether
we have data
etc
-
A 'data quality' file containing information
about every trial/stratum including details of missing or fudged categories
of information, eligibility
etc
-
A 'balance' file marking the ratios of randomised
comparisons that are not 1 : 1
-
An 'update' file indicating the vintage of
the data (and noting anything currently being processed) for each trial/stratum
-
For each centre/group, a 'project' file containing
contact names, addresses, telephone and fax numbers and E-mail addresses
(including full previous details and persons no longer 'active'); general
information; specific information about each trial/stratum separately;
literature references; a complete calendar of past actions, correspondence,
data processing etc and a complete calendar of future actions required
(this could be used as a project planner, although in practice only Jon
Godwin sees it!)
-
A look-up file giving the 'brief' name for
each trial/stratum as it is to appear on forest plots
-
A look-up file giving the 'year code' for
each trial/stratum as it is to appear on forest plots
-
For each centre/group, an analysis-generated
file containing the results of data consistency and error checking for
each trial/stratum
-
For each centre/group, an analysis-generated
file containing the results of balance and distribution checking for each
trial/stratum
-
For each overview question, a 'steering file'
containing a list of all the relevant randomised treatment comparisons
-
A 'master list' of names, addresses, telephone
and fax numbers and E-mail addresses for everyone involved in the overview.
This is a superset of the people in the centre/group 'project' files (whose
details also appear, redundantly but with enhancements, in those files)
-
A 'data specification' file describing the
'standard format' for individual patient data
In many cases there are also 'frozen' versions
of the overview data for retrospective analysis.
There are, of course, numerous other files
'off the critical path' including lists of all treatment agents, 'guesses'
for missing values of prognostic indicators (not used) and many more experimental
things.
The files are all 'flat' human-readable
text; however, they conform to certain rules that enable them all to be
read by the analysis software too. Techies please note that the 'flat'
files are shadowed by sorted binary versions of the information for faster
processing - these are automatically regenerated whenever the corresponding
'flat' file is found to have changed.
The analysis software and detailed descriptions
of the files can be found via the Contents
List, in GALAXY Help and elsewhere.
Checking data
NOW
The following remarks are tailored to the
BC 2000 and BC 2005 overviews, but apply more or less equally to all of
the others.
There are three vital GALAXY programs for
data checking:
STARE - finds all the missing
items and errors
INQUIST - tabulates all the variables
and does statistical tests
LIFE - draws curves
When you have a new file to check, name it
BC:##.NEW (where ## is the centre/group number). There are already files
BC:##.DAT, which are currently the 'translated' stuff from BC 1995. This
doesn't matter, because a .NEW file is given priority over a .DAT file
by all the programs.
How to use the programs
Huh? STARE ##
- produces STA##.DAT, a list of
all the patient records with problems, with a summary at the end.
STA## is often HUGE and it's usually better
just to look at it on the screen. What we used to do was to edit it so
that only a manageable number of key problem records remained, and send
that to the trialist. (To print it: BURGER STA## H12)
Huh? INQUIST ##
- produces INQ##X1.DAT, a tabulation
of all the variables, and INQ##X2.DAT, the statistical tests.
We usually sent INQ##X1 to the trialist.
(To print the files: BURGER INQ##X1 H12 or BURGER INQ##X2 H12)
Huh? LIFE #### -1 1
- produces L####V.DAT, a graphics
file containing the life-table, accrual and follow-up curves for trial/stratum
####.
(To print the pictures: LASER L####V)
Note: STARE also creates a summary file BC:DBA##.DAT
and INQUIST creates BC:DBC##.DAT - these are useful for a quick look, and
are used by other programs as data quality indicators and to see how the
overview is progressing etc.
There is a fourth program LBASIC which
produces the lists of patients with overdue follow-up, missing death causes
etc
etc, but all this has probably been changed since the last overview.
There are also PC versions of STARE, INQUIST,
LIFE, BURGER, LASER, LBASIC etc if required.
[Return to the Contents
List]
[End of document, updated to 13 May 2005]