Hello, First good year with a new version of EpiData Analysis. Thank you for your work. I tested some commands from the first test version: - I guess ": =" instead of "=" is because of the Mac or Linux? In any case, it is complicated to teach (less intuitive). Except for the equivalences (like Stata with the "=" and the "==" ... - Same for "!" instead "/" ? - Please use both upper and lower case for commands and variable names - After the "freq" command, the percentages are written in the vertical form and should be followed by "! c" instead of "! r" - Some personal demands for pedagogical purposes: = The "means"! t "command should show the result of the" t "test in case of two averages, and the result of ANOVA in case of more than two averages. I know that que le "F" est the square of the "t" for comparison of two averages. = Provide a complement to the "regress" command to get the regression line (as in the old Epi-Info "(eg"! Line ") = For the comparison of percentages, in case of insufficient calculated numbers (Ncell <5), do not calculate the chi-square and the "p" of the chi-square, and present the Fisher test for two percentages (the software "R" also presents the Fisher test for more than two percentages). Good luck. Thank you to excuse my bad english....
Bernard BRANGER 11 bis, rue Gabriel LUNEAU 44000 NANTES FRANCE Mobile 33 (0)6 32 70 33 80 Courriel : branger-bono@wanadoo.fr
Some thoughts on Bernard’s comments. Ease of use in a teaching setting is important, but I appreciate why some of the decisions have been made (largely to simplify the parser).
On Jan 3, 2017, Bernard wrote:
- I guess ": =" instead of "=" is because of the Mac or Linux? In any case, it is complicated to teach (less intuitive). Except for the equivalences (like Stata with the "=" and the "==" …
I can live with this; I’ve not done much programming in a language that requires :=, but a few errors and I’m used to it.
- Same for "!" instead "/“ ?
! is hard to type; I would also prefer /. I think the only potential conflict is with arithmetic in a statement, but is there a case where statement with / for divide could also have an option on it?
- Please use both upper and lower case for commands and variable names
I think the proposal is to let the user use case for ease of reading, but that there should be no reason to have two variables Age and age. I don’t like mixed case for programming as it multiplies syntax errors greatly.
- After the "freq" command, the percentages are written in the vertical form and should be followed by "! c" instead of "! r”
Agree, these are column percentages; for consistency with ‘tables’, !r and !c should have the same meaning
- Some personal demands for pedagogical purposes:
= The "means"! t "command should show the result of the" t "test in case of two averages, and the result of ANOVA in case of more than two averages. I know that que le "F" est the square of the "t" for comparison of two averages.
There may be times when F is useful, even for comparing two means. Can we have both or option for both?
= Provide a complement to the "regress" command to get the regression line (as in the old Epi-Info "(eg"! Line ") = For the comparison of percentages, in case of insufficient calculated numbers (Ncell <5), do not calculate the chi-square and the "p" of the chi-square, and present the Fisher test for two percentages (the software "R" also presents the Fisher test for more than two percentages).
I liked the traditional Analysis option for exact tests, but also realize that the developers can make logical decisions here. Getting different output may be confusing to the new user. I wonder if there could be a ‘newbie’ setting that would provide more explanation of output to the statistically naive user. What I have never liked about SPSS is that is was so easy to get garbage out; it is more difficult to make SAS or R or Stata work if you don’t know what you are doing. Analysis should be somewhere in between; ease of use and relevant output have always been a strong point of Analysis.
Jamie
Dear list.
I will try to elaborate a little on our choices so far.
As Jens Lauritsen wrote in the release email, the choices are not final yet, and we are definitely open to changes following feedback.
Some aspects are very easy to answer:
1. At this point do not expect all statistics and graphs to be completed. This will follow later. 2. Specific aspects of a given command , e.g. F or T statistics for Means can be discussed later. 3. We will implement case such that age, AGE, Age are all reference to the same variable. 4. We will attempt to have the same function in all of EntryClient, Manager and Analysis for the same feature (e.g. the browse window).
We are in particular open to discuss these aspects at this point:
A. Which sign should be used (= vs :=, <> vs !=, == vs =, etc) An aspect of this is which signs are easy to make on keyboards in all countries, e.g. ! is easy on the danish keyboard
B: How should we define select - as in current Analysis v2, where an additional select is incremental adding to a previous select. Or as in current prototype, where a given select only works for one command at a time or on a begin ... end; block
C: How should we define options. E.g as in Analysis v2 with a symbol (which cannot be "/"), but it could also as (notice the "," and no symbol for each option): /command/ /variable [variable ... ]/ , /option/ [ := </expression/> ] /options/ [...]
And for testing we would like your feedback: * Syntax: Is the current syntax reasonably easy to understand and use.
* Correctness: Are the calculations correct, this DOES include the "means" and freq commands, but is mostly related to expressions, functions and conversions of values.
* Usage: Do the basic commands work (eg. read/save/use) . Compared to the old Analysis we have tried to group common functionality into "families" - eg. when you wish to change properties of a variable/valuelabel/dataset then it is always the "edit" command that is use.
Please be careful by looking into the help system - see the help menu. As well the F1 button short example as the "commands.html" file that is included in the .zip file when downloading the program. This is complete list of commands and functions that is currently implemented. We will keep updating this list as we extend the program with more functionality.
Kind regards, Torsten Bonde Christiansen EpiData
On 2017-01-03 16:09, EpiData development and support wrote:
Some thoughts on Bernard’s comments. Ease of use in a teaching setting is important, but I appreciate why some of the decisions have been made (largely to simplify the parser).
On Jan 3, 2017, Bernard wrote:
- I guess ": =" instead of "=" is because of the Mac or Linux? In any case, it is complicated to teach (less intuitive). Except for the equivalences (like Stata with the "=" and the "==" …
I can live with this; I’ve not done much programming in a language that requires :=, but a few errors and I’m used to it.
- Same for "!" instead "/“ ?
! is hard to type; I would also prefer /. I think the only potential conflict is with arithmetic in a statement, but is there a case where statement with / for divide could also have an option on it?
- Please use both upper and lower case for commands and variable names
I think the proposal is to let the user use case for ease of reading, but that there should be no reason to have two variables Age and age. I don’t like mixed case for programming as it multiplies syntax errors greatly.
- After the "freq" command, the percentages are written in the vertical form and should be followed by "! c" instead of "! r”
Agree, these are column percentages; for consistency with ‘tables’, !r and !c should have the same meaning
- Some personal demands for pedagogical purposes:
= The "means"! t "command should show the result of the" t "test in case of two averages, and the result of ANOVA in case of more than two averages. I know that que le "F" est the square of the "t" for comparison of two averages.
There may be times when F is useful, even for comparing two means. Can we have both or option for both?
= Provide a complement to the "regress" command to get the regression line (as in the old Epi-Info "(eg"! Line ") = For the comparison of percentages, in case of insufficient calculated numbers (Ncell <5), do not calculate the chi-square and the "p" of the chi-square, and present the Fisher test for two percentages (the software "R" also presents the Fisher test for more than two percentages).
I liked the traditional Analysis option for exact tests, but also realize that the developers can make logical decisions here. Getting different output may be confusing to the new user. I wonder if there could be a ‘newbie’ setting that would provide more explanation of output to the statistically naive user. What I have never liked about SPSS is that is was so easy to get garbage out; it is more difficult to make SAS or R or Stata work if you don’t know what you are doing. Analysis should be somewhere in between; ease of use and relevant output have always been a strong point of Analysis.
Jamie
EpiData-list mailing list EpiData-list@lists.umanitoba.ca http://lists.umanitoba.ca/mailman/listinfo/epidata-list
Thanks Torsten. Here are some comments. I hope others will respond as well. Jamie
On Jan 4, 2017, at 07:17, EpiData development and support epidata-list@lists.umanitoba.ca wrote:
A. Which sign should be used (= vs :=, <> vs !=, == vs =, etc)
My preference for logical comparisons: ==, <> or !=
An aspect of this is which signs are easy to make on keyboards in all countries, e.g. ! is easy on the danish keyboard
B: How should we define select - as in current Analysis v2, where an additional select is incremental adding to a previous select. Or as in current prototype, where a given select only works for one command at a time or on a begin ... end; block
I prefer the begin … end blocks.
I would like to see:
select <boolean> begin <statements>; end; else begin <statements>; end;
This would implement if-then-else functionality at the record level, leaving if-then-else for use with globals; by the way, currently if-then will compare the first record if a variable name is used.
C: How should we define options. E.g as in Analysis v2 with a symbol (which cannot be "/"), but it could also as (notice the "," and no symbol for each option): /command/ /variable [variable ... ]/ , /option/ [ := </expression/> ] /options/ […]
I like the following: select group==1 means age, by=sex t vl
And for testing we would like your feedback:
- Syntax: Is the current syntax reasonably easy to understand and use.
It seems that programs require ; at the end of a line. This will need to be emphasized (where to use ; and where not). The examples show this pretty well.
participants (1)
-
EpiData development and support