- EpiData-list - University of Manitoba Mailing Lists

Reserved words and a warning on using "if constructs"
by epidata-list＠lists.umanitoba.ca 04 Nov '05

04 Nov '05

1)A member commented on the use of date as a reserved word and the use of chk and dbf files. CHK files are only read with REC files. If you read CSV (delimited files) or DBF files no CHK file of the same name as the csv or dbf file will be used. If you wish to add labels (comment legal) for data in dbf or csv files then do: read myfile.dbf savedata xname * create a chk file with the name xname.chk read xname // obviously any name can be used xname is just an example On the topic of reserved names it is difficult because so many different languages exist. But in general do not use names of functions. E.g. avoid year, date and month use instead Y, d , m or vyear vdate vmonth 2. Using if the use of temporary select is a strong facility, e.g.: count if age > 50 tab sex agegrp if HbA1c > 6 tab sex agegrp if Country = 1 but currently DO NOT USE IF in combination with options for any command, e.g.: tab sex agegrp /Missing if Country = 1 this would be read as: tab sex agegrp /Missing Country = 1 meaning that all data was recoded for variable country. With next release this will be fixed. A new build is on the way fixing this and the other known (few) bugs. But I wish to finalize the aggregate (summary output statistics) command first. 3. Screen fonts in EpiData dataentry can be defined with options and a combination of the use of @ and blanks. See the help file. Screen fonts in analysis are defined with the "Set viewer font name", see the documentation on output formats installed with the programme. Regards Jens Lauritsen EpiData Association

1 0

EpiData editor, set tables
by epidata-list＠lists.umanitoba.ca 03 Nov '05

03 Nov '05

greetings. (1)the problem with the set command NOT working was because when the .DBF data set was imported and then saved as a .REC the 'new' data set contained the var. name DATE. EpiDataStat went crazy, sorta. no error message about reserved words or anything so helpful. so, it wasn't the .CHK constuct. (2) how does one control the EpiData editor? i spent considerable time creating a data collection instrument and the result was less than desirable when trying to print the form either at Data Entry or from the editor. so there must be something in the fine print which i haven't seen. how to get it to wrap at 80 cols., what about type size, etc. page length etc. thanks

1 0

set table value and label on/off and getting back to analysis output view
by epidata-list＠lists.umanitoba.ca 30 Oct '05

30 Oct '05

Set table value = off - and on Only works if the variables has a defined comment legal With the bromar.rec supplied with analysis: set table value = on tab sex sex No. % 1 Female 490 12.17 2 Male 3537 87.83 Total 4027 100% set table value = off tab sex sex No. % Female 490 12.17 Male 3537 87.83 Total 4027 100% set table label = off TABLE LABEL=OFF tab sex sex No. % 1 490 12.17 2 3537 87.83 Total 4027 100% (2) Analysis: when using Help: select (a) how to, (b) commands&func, (c) Intro when u select either of these u cannot get back to Analysis without first Quiting the program. To show output again Either: 1. Click on "Viewer" in tool bar and "Show Output" 2. Have your cursor active somewhere on the output (e.g. in the middle of help) and press ESC 3. Issue the command CLS (clear screen) and produce new output. The cls command is important every so often if you think the programme is slowing down. This is due to the way output is formed. regards Jens Lauritsen EpiData Association

1 0

How to: select with dates and an open question ? ?
by epidata-list＠lists.umanitoba.ca 30 Oct '05

30 Oct '05

This should work: * ---------------------------------------- DEFINE DATEMIN <dd/mm/yyyy> DEFINE DATEMAX <dd/mm/yyyy> LET DATEMIN = dmy(? First date: e.g. (02/12/1989) ?,"dmy") LET DATEMAX = ? Second date : e.g. (02/12/1999) ?,"dmy") SELECT (DATEADM <= DATEMIN) AND (DATEDIS >= DATEMAX) FREQ AGE ................................................................ or to select directly in comparison with a specific date using the dmy function: select date(? Write date, dd/mm/yyyy ?,"dmy") > dmy(10,02,1999) Jens Lauritsen EpiData Association

1 0

Analysis: Set, Help
by epidata-list＠lists.umanitoba.ca 30 Oct '05

30 Oct '05

greetings. (1)Analysis doc. says default value of SET TABLE VALUE = OFF My Experience is that the command does not work at all. It is ON all the time and cannot be turned OFF or ON using the set command. (2) Analysis: when using Help: select (a) how to, (b) commands&func, (c) Intro when u select either of these u cannot get back to Analysis without first Quiting the program. Suggest a CLOSE window, similar to Browse (unless i'm missing the return somewhere) -----Original Message----- From: epidata-list(a)lists.umanitoba.ca [mailto:epidata-list@lists.umanitoba.ca] Sent: Sat 10/29/2005 4:50 AM To: epidata-list(a)lists.umanitoba.ca Cc: Subject: [EpiData-list] TABLES OUTPUT greetings. i was going to send 2 .doc files demonstrating the following 'problem' but, they were more than 25k. 1. have database using .chk which contains a LABELBLOCK the LABELBLOCK was constructed 1-16 with names assigned to each(of coourse) 2. ran Analysis and issued the command TABLES SEX SITECODE 3. result was what appeared to be a random display of SITECODE using the LABELs assigned 4. it turned out to be the problem with the number ordering 1 10 11 12 13 14 15 16 2 3 4 5 6 7 8 9 --so, i suggest either an alphabetc listing of labels or in order it was created 1-16 etc. --when i removed the LABELBLOCK the output was 1-16 and it should be. also, i noticed that another person is having the same issue with the FREQ resulting in error 'Too Many Catagories" thanks, kelly

1 0

Bed occupancy - problems with date calculation...
by epidata-list＠lists.umanitoba.ca 29 Oct '05

29 Oct '05

Hi. I am trying to write a PGM in order to calculate the bed occupancy in a defined period of time. As dates can be converted to sequential integer numbers, I wrote the following, considering DATEADM (admission date), DATEDIS (discharge date) and AGE as database variables: * ---------------------------------------- DEFINE DATEMIN <dd/mm/yyyy> DEFINE DATEMAX <dd/mm/yyyy> LET DATEMIN = ? First date : ? LET DATEMAX = ? Second date : ? SELECT (DATEADM <= DATEMIN) AND (DATEDIS >= DATEMAX) FREQ AGE *----------------------------- After typing first date: 15/10/2005 (even "15/10/2005"), I get the error message: "Data type mismatch". If I type 15102005, I get "Invalid date", what sounds reasonable. But I can only entry the desired date when I type 75162, what is NOT the respective sequential number of days since 31/12/1899 ("EpiData handles dates as float-type numbers counting the number of days since 31/12/1899" as can be read in EpiData Help) What am I doing wrong? Thank you in advance. MARCO MOURA ================================== Dr. MARCO Antonio MOURA Reis, MD, PhD Discipline of Medical Informatics Department of Medicine University of Region of Joinville - UNIVILLE Joinville (SC) - BRAZIL E-mail: mmoura(a)infomedica.com.br ==================================

1 0

How to: How is sort order of tables defined ? (get tables in right order)
by epidata-list＠lists.umanitoba.ca 29 Oct '05

29 Oct '05

One way of getting other people to see problems is to have analysis creating random data with a combination of the command generate (which produces an empty data file) and the random functions. Then we do not need the attached files. The reason for NOT allowing attachments is that this is a very good way of protecting all on the list from virus spreading E.g. for the question from today run these commands: First: create a chk file with the labels: ....................... t.chk ................ labelblock label s 1 "London" 2 "Copenhagen" 3 "Madrid" 4 "New Delhi" 5 "Madras" 6 "Pretoria" 7 "Kiev" 8 "Hong Kong" 9 "Melbourne" 10 "Sao Paolo" 11 "Santiago" 12 "Ottawa" 13 "New Orleans" 14 "Bangkok" 15 "Honolulu" 16 "Havana" 0 "Oslo" 17 "Odense" 18 "Paris" 19 "Budapest" end end sitecode comment legal use s end sex comment legal 0 male 1 female end end ..... save the text above as t.chk ........................ ............ Then create a pgm file: t.pgm: ................................ * Example file for creation of test data: close * create an empty data set with 1000 observations: generate 1000 define sex # define sitecode ## define sitecode2 ## * now assign values to the variables (random numbers) sex = ran(2) // sex will get numbers 0 and 1 sitecode = random(20) // sitecode will get numbers from 0 to 19 sitecode2 = sitecode savedata t * Now we have saved a data file with 1000 numbers, * but we need to read it again to get the labels read t /close * turn on labels, but turn off numbers for the tables: set table label = on set table value = off tab sex sitecode tab sex sitecode2 * notice that the order is the same as the numerical code for both variables * now we also turn on the numbers for the sitecode: set table value = on tab sex sitecode ................................ end t.pgm When you run the t.pgm you will notice that the first table of sex and sitecode shows this: SEX SITECODE male % female % Total % Oslo 25 {52.1} 23 {47.9} 48 {100.0} London 26 {48.1} 28 {51.9} 54 {100.0} .... snip ....................... Whereas the last with the values turned on shows: set table value = on tab sex sitecode SEX SITECODE 0 male % 1 female % Total % 0 Oslo 25 {52.1} 23 {47.9} 48 {100.0} 1 London 26 {48.1} 28 {51.9} 54 {100.0} 10 Sao Paolo 26 {47.3} 29 {52.7} Which means that for crosstables there is apparently a sorting going on on the first column. Notice that the two variables sitecode and sitecode2 are of type numerical. If we instead had used string variables: define v2 __ v2 = string(sitecode) tab v2 * then the order of v2 would be sorted meaning the "0 oslo - 1 london - 2 sao paolo" sequence * but without the text. To remedy this the trick would be to say: v2 = " " + v2 if length(trim(v2)) < 2 Then a new "tab v2" would show the correct table ------------------------------------------ Lesson: Obviously more user control of sort order of cross tables is on the wish list. But the feature is not part of the current most wanted list from my point of view. We have been experimenting a bit with sorting, but it is not all that easy. By converting btw. variable types and turning values and labels on and off you can get a lot of control as it is now. A feature which is expected to be available later this year is to be able to take the text label from a variable and save in a string variable, e.g. v2string = label(sitecode) or v2string = string(sitecode) + label(sitecode) (this is not implemented now) An undocumented sorting feature is available - but will not be documented until I am sure how to control sorting. Users are welcome to use it. The table will be sorted in reverse numerical order. But the problem with labels and values above persist. table sex v2 /rd or tab sex v2 /Rowdes Jens Lauritsen EpiData Association > 1. have database using .chk which contains a LABELBLOCK > the LABELBLOCK was constructed 1-16 with names assigned to each(of coourse) > 2. ran Analysis and issued the command TABLES SEX SITECODE > 3. result was what appeared to be a random display of SITECODE using the LABELs assigned > 4. it turned out to be the problem with the number ordering > 1 > 10 > 11 > 12 > 13 > 14 > 15 > 16 > 2 > 3 > 4 > 5 > 6 > 7 > 8 > 9 > --so, i suggest either an alphabetc listing of labels or in order it was created 1-16 etc. > --when i removed the LABELBLOCK the output was 1-16 and it should be. also, i noticed that > another person is having the same issue with the FREQ resulting in error 'Too Many Catagories" > thanks, kelly > >

1 0

TABLES OUTPUT
by epidata-list＠lists.umanitoba.ca 29 Oct '05

29 Oct '05

greetings. i was going to send 2 .doc files demonstrating the following 'problem' but, they were more than 25k. 1. have database using .chk which contains a LABELBLOCK the LABELBLOCK was constructed 1-16 with names assigned to each(of coourse) 2. ran Analysis and issued the command TABLES SEX SITECODE 3. result was what appeared to be a random display of SITECODE using the LABELs assigned 4. it turned out to be the problem with the number ordering 1 10 11 12 13 14 15 16 2 3 4 5 6 7 8 9 --so, i suggest either an alphabetc listing of labels or in order it was created 1-16 etc. --when i removed the LABELBLOCK the output was 1-16 and it should be. also, i noticed that another person is having the same issue with the FREQ resulting in error 'Too Many Catagories" thanks, kelly

1 0

hi Jamie
by epidata-list＠lists.umanitoba.ca 29 Oct '05

29 Oct '05

please send email address.... keb1(a)cdc.gov -----Original Message----- From: epidata-list(a)lists.umanitoba.ca [mailto:epidata-list@lists.umanitoba.ca] Sent: Wed 10/26/2005 6:26 PM To: epidata-list(a)lists.umanitoba.ca Cc: Subject: [EpiData-list] how do you get the file to test the problem? Use EpiData to archive the file - it will be a fraction of the original size. You an do this from the Tools menu. If still too large, then we'll have to exchange e-mail addresses off the list somehow Jamie Hockin Public Health Agency of Canada

1 0

Administrative comments on recent issues on EpiData-List
by epidata-list＠lists.umanitoba.ca 27 Oct '05

27 Oct '05

Good colleagues. A few administrative issues: 1. The list has been set up to enhance communication by e-mail and to limit size of each posting. For users with fast internet size of a mail is not a concern, but for people on slow modem lines large mails creates large problems. So as of now the following rules guide the list: a. Attachments must be small b. Text in e-mail must be small. A combined limit of 25kb has been defined. A few days back a user attached a rec file with the size of almost 1mb which is far exceeding this size. Individual users can exchange large attachments outside the list directly as suggested by Jamie Hockin. 2. Only certain attachments are allowed, e.g. small rec, chk and qes files. 3. Minimise the number of lines when you answer or comment on other mails. If a user asks a question - then on the reply only keep the minimal part indicating the topic of the question and reply. Otherwise the complete originating mail can be repeated several times. 4. Bouncing. When a person has an automatic mail saying e.g. "I am out of office please ....." then when a notice from the list comes to that person a message is sent to the person. The automatic message is sent back to the list, which then replies to the user etc. So if no mechanism controls several hundred mails can bounce back and forth. Therefore some users will experience to be automatically signed off the list. This is not a censorship or similar but a mechanism to avoid that everyone gets many "I am out of office..." messages. If you are excluded by "bounce-control" then just sign on to the list again when you are back from the leave. Or maybe create a different e-mail account which you use for receiving list messages. 5. Write in a specific and short manner and protect your e-mail adress You should be aware that messages you send to the list are: - distributed to all members (currently nearing 200) - is kept as part of the archive and is available for searching on internet. So to minimise the chance that your e-mail is trapped and used in a virus spread do not write it directly if you wish to receive personal mails, but write something like xxx at yyyy zz or similar. instead of xxx(a)yyyy.zz 6. Only users who are members of the list can submit messages. The first message by a new list member is sent to list administrators for approval. We could turn this off, but it will result in advertisements or spam messages every so often. I assume list members appreciate the principles above, but I am willing to discuss some of them. E.g. the size of allowed attachments and mails. Kind regards Jens Lauritsen EpiData Association

1 0