E N D
1. Stata - Experiences and Needs of a Professional Services Organisation Dr. Andreas Stiehler
Berlecon Research GmbH
Oranienburger Str. 32
10117 Berlin
as@berlecon.de
Fon/Fax: +49 30 285296-0/-29
2. Intro: Berlecon Research
Requirements for statistic software
Deploying Stata 8: main challenges
Conclusions and wish list
3. I. Intro: Berlecon Research GmbH We analyze the potential of new technologies.
Analyses are typically published as reports
Our clients are demanding and have high standards.
IT executives from large and medium-sized companies
Technology and consulting companies
Public institutions (EU commission, BMWA etc.)
Empirical work makes up a significant part of the research.
About 3 - 5 surveys per year (increasing importance)
Data typically delivered by market research companies
4. II. Requirements for statistics software
5. II. Requirements for statistics software We need less sophisticated statistics
> 90% binomial and ordinal data
mainly descriptive statistics needed (tabulate, tabsum)
but high quality graphs and tables!
layout of the reports is a competitive factor
We only have a limited amount of time
for both exploring software opportunities and computing
but need highly reliable data!
errors in data analysis would seriously damage our reputation
6. II. Requirements for statistics software July 2004: Start of internal project with two objectives:
Choosing a new statistics software and - based on it -
Finding a way to streamline the creation of tables and graphs
Decision in favour of Stata 8, decisive features are:
Programming opportunities: allows for customisation of outputs and exploiting economies of scale
Intuitive user interface: facilitates the exploration of possibilities that Stata offers.
Graphs: high quality, no additional software needed
Further features: Statalist, Mac support, price
7. III. Tables and Graphs: main challenges
8. III. Tables : Main Challenges Goal: Preparation of data for custom tables
Overview tables typically compile results from several tabulate(s)
Often further transformation of results needed (e.g., summary data for general approval = fully agree + somewhat agree).
9. III. Tables : Main Challenges Solution today: Processing of data via matrices
Tabulate results to matrix: tab A B, col nofreq matcell (AB)
Transformation of data using matrix operations: percentages, aggregation, compiling for the presentation in table.
Matrix to data table: svmat (AB), names (newsol1)
Copying of table-ready data to Excel or Word
Advantages
Significant time savings (half of the time needed compared to transfer of results on a case-by-case basis)
Error reduction (copying only once)
10. matrix def e = 1,1,1,1,1foreach var of varlist q31* {/*********Country******************************/tabulate `var' country, matcell(c`var')matrix define gesc`var'= e * c`var'matrix define c`var' = c`var' * inv(diag(gesc`var')) * 100/***********Education*************************/tabulate `var' edu, matcell(e`var')matrix define gese`var'= e * e`var'matrix define e`var' = e`var' * inv(diag(gese`var')) * 100/***************All together***************************/matrix define `var' = c`var''\e`var'matrix define `var' = `var'* (1,1,0,0,0)'svmat `var', names (z`var')foreach var of varlist z`var'* {replace `var' = round(`var') }}
11. III. Graphs: Main challenges Graphs at Berlecon:
About 7 different graph types in use
Customisation time: about 2 weeks (GUI -Submit - big help)
Use of program files makes it economic (re-use of do.files)
Extensive use of overlaid two-way charts
12. IV. Conclusions and wish list
13. IV. Conclusions Stata has the potential to serve as a helpful tool for the specific needs of professional services organisations.
Ability to streamlined and less error prone generation of table-ready data.
High quality graphs - no additional software needed.
Intuitive user interface and collaboration via Statalist facilitates programming for non-experts.
There are several challenges:
Standard output typically not sufficient for further processing in tables and graphs.
Finding appropriate solutions is challenging - but its worth it.
14. IV. Wish list Tables:
Option to:
Design tables within the Stata program!!!
Save percentages in matrices
Display zero rows/columns to matrices in case of empty cells
Graphs:
Option to define colours for value labels in stacked bar charts as standard (as it is possible for pie charts)
Text options for labels on y/x-axis (including line breaks, size
)
Book A visual guide to Stata for non-standard graphs
15. /***Tabulate results to data editor*********/matrix define e7 = 1,1,1,1,1,1,1foreach var of varlist f7* {tabulate `var' [fweight=rweight], matcell(`var')matrix define ges`var' = e7 * `var'matrix define `var' = `var' * inv(ges`var')* 100}matrix define y7 = f7a,f7b,f7cmatrix define y7 = y7'svmat y7/*Stack bars*/gen y7stack7 = y71+y72+y73+y74+y75+y76+y77gen y7stack6 = y71+y72+y73+y74+y75+y76gen y7stack5 = y71+y72+y73+y74+y75gen y7stack4 = y71+y72+y73+y74gen y7stack3 = y71+y72+y73gen y7stack2 = y71+y72gen y7stack1 = y71}
16. twoway ///(bar y7stack7 zaehl31, horizontal blcolor(black) bfcolor(gs13) barwidth(0.5)) /// (scatter zaehl31 y7scat7, msymbol(none) mlabel(y77) mlabsize(medsmall) ///mlabcolor(black) mlabposition(0)) //////(bar y7stack6 zaehl31, horizontal blcolor(black) bfcolor(gs3) barwidth(0.5)) /// (scatter zaehl31 y7scat6, msymbol(none) mlabel(y76) mlabsize(medsmall) ///mlabcolor(white) mlabposition(0)) //////(bar y7stack5 zaehl31, horizontal blcolor(black) bfcolor("204 255 204") barwidth(0.5)) /// (scatter zaehl31 y7scat5, msymbol(none) mlabel(y75) mlabsize(medsmall) ///mlabcolor(black) mlabposition(0)) //////(bar y7stack4 zaehl31, horizontal blcolor(black) bfcolor("0 204 0") barwidth(0.5)) ///(scatter zaehl31 y7scat4 .........
17. ..../// , ylabel( ///0.85 "Cost " ///1.15 "leadership " ///1.4 "..........................." ///1.85 "Quality " ///2.15 "leadership " ///2.4 "..........................." ///2.85 "Innovation " ///3.15 "leadership " ///3.4 "..........................." ///3.85 "Market " ///4.15 "leadership " ///4.4 "..........................." ///4.85 "Vertical " ///5.15 "focus " ///5.4 "..........................." ///5.85 "Focus on " ///6.16 "enterprise size" ///, .....