behavioral data management using visual basic for applications to automate data capture and analysis

Behavioral data management using VISUAL BASIC FOR APPLICATIONS

to automate data capture and analysis

Maurice L. Sipos a,*, Richard E. Sweeney b

a Advanced Assessment Branch, Drug Assessment Division, United States Army Medical Research Institute of Chemical Defense, Aberdeen Proving

Ground, MD 21010, USAb RESECO Research Engineering Consultants, PO Box 2311, Upper Darby, PA 19082, USA

Received 23 September 2002; received in revised form 14 May 2003; accepted 14 May 2003

Journal of Neuroscience Methods 128 (2003) 53�/65

www.elsevier.com/locate/jneumeth

Abstract

Many researchers are familiar with the spreadsheet capabilities of Microsoft Excel†, but have never explored using customized

VISUAL BASIC FOR APPLICATIONS (VBA) macros embedded in the program. At the United States Army Medical Research Institute

of Chemical Defense (USAMRICD), the implementation of VBA program code to carry out repetitive operations has resulted in a

tremendous savings in both the time and manpower required to reliably capture, analyze, and plot data from research protocols. A

set of ‘‘template’’ workbooks was developed and is used to organize data from different types of studies. At the heart of the concept

is a ‘‘setup information’’ worksheet onto which the user enters information about the study (i.e. the number of subjects, groups,

graphs). Clicking a control button on this worksheet launches the VBA code that creates worksheets for each subject, group and

chart specified. A ‘‘pairmatch’’ function allows the user to create groups either randomly or by pair matching based on user-

specified variables. Controls are provided that run VBA program code to parse subject data files into the proper worksheets and

perform group averages. Charts are updated automatically whenever group averages change so that the researcher always has an up-

to-date plot available. The ability to reformat the captured data onto a ‘‘statistical output’’ worksheet allows data to be imported

easily to statistical software packages. The concept is being used for several different types of studies at USAMRICD and has

significantly reduced not only amount of time spent on data management, but also the number of data entry errors. Readers

interested in acquiring an electronic copy of the startle workbook example, which contains the complete VBA code, should send the

request to the authors at [email protected] or [email protected].

# 2003 Elsevier B.V. All rights reserved.

Keywords: Data entry; Data management; Automation; VISUAL BASIC FOR APPLICATIONS; Macros; Microsoft Excel

1. Introduction

Studies examining the effects of repeated exposures to

low doses of chemical warfare nerve agents at the

United States Army Medical Research Institute of

Chemical Defense (USAMRICD) typically involve the

daily collection of several physiological or behavioral

measurements from subjects over several weeks or

months (Fig. 1 illustrates the time course of a typical

study). These studies generate copious amounts of data

that must be collected, organized, stored, compiled and

analyzed. Prior to the inception of this automated

system, skilled technicians working on the study were

forced to process the data at least partially by hand,

which often required them to devote a significant

portion of their time to routine tasks such as data entry,

computations, or file manipulations. Manual data

management typically involved reorganizing the data

files produced by laboratory software packages and

importing them into a spreadsheet. Subsequently, in-

dividual subject data could be imported into a statistical

package for data analysis and graphical presentation or

could be graphed directly using the spreadsheet pro-

gram. In addition to consuming an inordinate amount

* Corresponding author. Present address: Army Physical Fitness

Research Institute, US Army War College, 122 Forbes Avenue,

Carlisle, PA 17013, USA. Tel.: �/1-717-245-4580; fax: �/1-717-245-

4558.

E-mail address: [email protected] (M.L. Sipos).

0165-0270/03/$ - see front matter # 2003 Elsevier B.V. All rights reserved.

doi:10.1016/S0165-0270(03)00150-X

mailto:[email protected]

of time, processing data manually was error prone and

required additional time to ensure the accuracy of data

entry.

In the present paper, the benefit of using VISUAL

BASIC FOR APPLICATIONS (VBA) macros to facilitate

data management and analysis is discussed. Microsoft

Excel† was chosen as the platform around which the

automated processing system would be centered.

Although many individuals are already familiar with

the EXCEL program, most have never explored the

possibility of using VBA macros embedded within the

application to perform routine tasks. The present manu-

script targets individuals with an intermediate level of

expertise: individuals who know what VBA macros can

do, but have never considered applying them to his/her

own laboratory. More inexperienced users are directed

to review more general references to gain a basic

understanding of what VBA is and how macros can be

used to automate repetitive functions (i.e. Jacobson,

1999; Wells and Harshbarger, 1997, or the VBA help

files).

Using VBA macros to perform repetitive data manip-

ulations increased productivity and reduced errors in

data entry. The macros were embedded into the work-

sheets of an EXCEL workbook that also served to

organize, analyze, plot and store the data. A workbook

was created as a template that could be copied andrenamed for use during studies examining the effects of

chemical warfare nerve agents on the acoustic startle

response in guinea pigs. Although this paper is based on

using VBA macros to process acoustic startle response

data, VBA macros have also been used at USAMRICD

to process data collected from active avoidance studies,

radial arm maze studies, and operant procedures, as well

as to process data collected using electronic laboratorymonitoring systems that collect body weight data and

temperature. A copy of the startle workbook template,

which contains the complete VBA code, may be

obtained from [email protected] or maurice.sipos@

us.army.mil.

2. Methods and results

2.1. Building the workbook

Since many parameters can change from study to

study, a ‘‘Setup Information’’ worksheet was included in

Fig. 1. Generalized time line for a study compatible with this software. Note that the baseline, challenge, and recovery periods can be of arbitrary

duration.

M.L. Sipos, R.E. Sweeney / Journal of Neuroscience Methods 128 (2003) 53�/6554

Fig. 2. The macros are contained in a ‘‘Template’’ workbook that contains worksheets from which Subject, Group, Statistics, and Charts are copied.

In this particular example, 60 subjects were assigned to one of ten groups based on treatment (saline or 0.4�/LD50 VX) and food availability (Ad

libitum, 100% of the recommended daily requirement (RDR), 85% of the RDR, 75% of the RDR, or 65% of the RDR).

Fig. 3. The subjects are randomly assigned into treatment groups and listed into columns ‘‘A’’ through ‘‘G.’’ See Fig. 2 legend for the details of the

study included in this example.

M.L. Sipos, R.E. Sweeney / Journal of Neuroscience Methods 128 (2003) 53�/65 55

the template workbook that would allow users to

customize the program to meet their individual needs

(see Fig. 2). On this worksheet, the user must enter the

total number of subjects in the study, the identification

number for the first subject, the number of study groups

and treatment groups included in the study, group labels

and their corresponding graph colors, the total number

of charts to be created, the names, titles, and specific

information for each chart, the directory from which

unprocessed data files are retrieved, and the directory

into which processed data files should be deposited.

Once the required information has been entered, the

user clicks on the ‘‘Create Subject Sheets’’ button to

initiate a VBA macro that directs the analysis workbook

to ‘‘build’’ itself based upon the setup information.

Although it has not been incorporated into the startle

template workbook, the VBA code could be used to

check the user’s entries for validity (based on rules set up

in advance). Annotated excerpts of the VBA code that

performs the ‘‘self-build’’ task are provided in section 1

of Appendix A. The self-build feature saves an extra-

ordinary amount of time when compared with manually

interfacing with the EXCEL program to create the charts

and groups. During building, many worksheets are

created in the workbook as described below.

1) A subject data worksheet is created from the

‘‘Subject X’’ template worksheet for each subject

in the study and is named according to the subject’s

identification number.

2) A group data worksheet is created from the ‘‘Group

X’’ template worksheet for each group defined in

the study and is named appropriately.3) A chart type worksheet is created from the ‘‘Chart

X’’ template sheet for each requested chart and is

named appropriately.

The user must create these template sheets manually

depending on the information that is collected during

the study. The organization used for the subject sheet in

a sample acoustic startle response study is shown in Fig.

5. It was useful to devote one row for each startle session

and single columns for date, group, and session identi-

fiers and for each variable measured during the session.

The organization of the group template can be similar to

the subject sheet, but would contain averages of results

across the subjects within the group. The chart template

sheet contains a single chart that plots one data series

for each group in the study. The y-source data for each

series is taken from a measurement column on a group

worksheet. The x-source data is taken from the corre-

sponding date column on the group worksheets. As a

result, each chart is devoted to a single variable

measured in the study and facilitates the comparison

of daily group means by dependent variable during the

course of the study. To create the chart, the ‘‘Group X’’

template was filled with temporary data to provide a

target for the chart’s series objects. During the work-

book build operation, the source data references for all

data series in each chart are automatically calculatedand pointed to appropriate cells on the group sheets.

Means, standard errors of the mean, and colors of the

lines for each group on each chart are entered auto-

matically under program control. After creation, the

charts always display the values (means and standard

errors of the mean) that are present in the group

worksheets.

2.2. The ‘‘Pairmatch’’ worksheet

As the workbook is built, information about the study

is entered automatically onto the ‘‘Pairmatch’’ work-

sheet. Fig. 3 shows a sample ‘‘Pairmatch’’ worksheet for

a new workbook before data are processed. Subject

identifiers are listed in columns ‘‘A’’ and ‘‘F.’’ Althoughnot used in the acoustic startle response sample study,

two columns are reserved for subject identifiers for cases

in which subjects can be identified using both a name

and a number. For example, nonhuman primates are

often given a number and a name (i.e. 89�/154 and

‘‘Alpha,’’ respectively). Columns ‘‘B’’ and ‘‘C’’ in Fig. 3

display empty cells for each animal, but these are

replaced by the average startle response values to 100and 120 dB stimuli collected over the most recent 5 days

of data collection prior to pair matching. Column ‘‘G’’

lists the treatment group to which each animal has been

randomly assigned by the macro. Columns ‘‘I’’ through

‘‘O’’ display group averages of the measurements used

for pair matching. This information is not valid until

pair matching, which will be discussed shortly, has been

performed. The information in cells ‘‘V2’’ through‘‘X16’’ is used for creation of an output worksheet

that presents data in a format that can be easily

exported to the SigmaStat† statistical package.

After the workbook has been built, the VBA macro

automatically renames the template to prevent the user

from inadvertently overwriting the original. The ‘‘Pair-

match’’ worksheet contains several control buttons that

run VBA macros that enable users to perform routinedaily tasks. Although data from multiple days can be

processed together, the power of the macros is that they

enable the user to examine the data immediately to

determine what effects are occurring on a daily basis. To

process data, a technician executes the steps outlined

below:

1) Move the raw data files (produced by the laboratory

equipment) for the day into the ‘‘input’’ directory

area specified in cell ‘‘B3’’ of the ‘‘Setup Informa-tion’’ worksheet.

2) Click the ‘‘Crunch New Data’’ button to launch a

macro that automatically reads each raw data file,


distributes its data to the appropriate cells on the

subject worksheets, and moves each file to the

‘‘archive’’ directory specified in cell ‘‘B6’’ of the

‘‘Setup Information’’ worksheet. Annotated ex-cerpts of the VBA code that performs these tasks

for the data files produced by the startle system are

provided in section 3 of Appendix A.

3) After 5 days of baseline data have been collected,

the user must click the ‘‘Do Pair Matching’’ button

to pair match the subjects in the study into equal

groups based on the variables specified in columns

‘‘B’’ and ‘‘C’’.4) The user can also update the graphs daily by

clicking the ‘‘Do Group Averages’’ button to launch

the macro that computes daily averages across

subjects for each group and updates the ‘‘Group’’

worksheets. This function is only enabled after pair

matching has been completed. Whenever the group

worksheets are updated, the charts linked to those

sheets are updated automatically.

2.3. Pair matching the subjects into groups

After the baseline period, information about each

animal’s baseline responses is available in the subject’s

worksheet. Worksheet formulas provide each subject’s

average response (averaged over 5 days) for each of the

two measures displayed in columns ‘‘B’’ and ‘‘C’’ of the‘‘Pairmatch’’ worksheet. Column ‘‘D’’ gives the average

of the two values for a subject and is used in the pair

matching process. Since variability exists in the strength

of individual responses to startle eliciting stimuli,

animals are pair matched to create individual treatment

groups that start out with similar group means and

variances. Pair matching is performed by clicking the

‘‘Do Pair Matching’’ button. In addition, this processsystematically assigns animals to behavioral test cham-

bers to assure that all treatment conditions are repre-

sented as equally as possible across test chambers.

Programmatically, pair matching was accomplished by

rank ordering the subjects based upon the magnitude of

their acoustic startle responses to 100 and 120 dB

stimuli. Subjects were then pulled by rank order and

assigned to study groups in a rotating fashion. Eachgroup is thereby assigned similar numbers of strong and

weak responders, balancing both group means and

variances.

Fig. 4 shows the ‘‘Pairmatch’’ worksheet after the pair

matching operation has been completed and the subjects

have been assigned into study groups. The information

in column ‘‘M’’ can be used to verify that the two study

groups within each treatment group in this examplehave similar group means. Pair matching can be done

only once, and the button is disabled after the procedure

has been performed.

2.4. Group averaging and statistical output

Any time after pair matching has been completed, the

‘‘Do Group Averages’’ button can be clicked to launch amacro that performs averages across all subjects within

a group for each day of the study. The ‘‘Group’’

worksheets are updated by this operation. Since the

charts are linked to the group worksheets, they are also

updated automatically. The format of the group work-

sheets is similar to the subject worksheet shown in Fig.

5. A sample chart worksheet is shown in Fig. 6.

At the end of the study, the user can click the ‘‘CreateStats Pages’’ button to launch a macro that copies the

‘‘STATS X’’ worksheet to a new worksheet named

according to the title given in ‘‘column ‘X’ in the

Pairmatch sheet’’. The information on the ‘‘Pairmatch’’

worksheet in cell ‘‘W4’’ is used to define the number of

‘‘STATS’’ worksheets to be created. There are 12

statistics titles listed in ‘‘X5 to X16’’ in this example,

but the figure in cell ‘‘W4’’ is eight. Thus, only the firsteight statistics sheets are created. The appropriate

dependent variable column on ‘‘Subject’’ worksheets

to be used in generating each statistics summary work-

sheet is defined by the list of numbers in cells ‘‘W5 to

W16’’. Each ‘‘STATS’’ output worksheet is named

based on the list of labels in cells ‘‘X5 to X16’’. Each

statistical output worksheet contains all the data in the

study for a single dependent variable (defined by thedata column number entered into the list starting at

‘‘W5’’). The subject, group and treatment identifiers as

well as the date values are also included in the statistical

output worksheets. A sample statistical output work-

sheet is shown in Fig. 7.

3. Discussion

The Microsoft Excel† component of the Office Suite

must be installed on the processing computer to use

VBA macros to organize and process data in the manner

described in this paper. This approach is most useful

when there are numerous data files generated by

laboratory equipment (in a standardized fixed format)

and when the data processing requires repetitive, time

consuming operations to retrieve individual subject datafrom these files, organize the data, perform descriptive

statistics and graph the data, export the data to a

statistics package, and store the original data files in an

organized fashion. Because the approach uses specia-

lized code to read the data file format output by

laboratory software, the addition of new file formats

to be read will require programming skills in the VBA

language to write the EXCEL macro programs to processdata files. We have successfully written templates

containing the proper code to manipulate data files for

the Hamilton�/Kinder Acoustic Startle System; the San


Fig. 4. As the study progresses through the baseline period, the daily raw data files are manually loaded into the ‘‘Unprocessed Data’’ directory and

the ‘‘Crunch New Data’’ button is pressed to launch a VBA macro that reads each raw data file; processes the file and distributes the results into the

subject worksheets; and optionally moves the data file to the ‘‘Processed Data’’ directory. A running average is kept for each subject’s ‘‘response’’

measured by a selected variable. At the end of the baseline period, the ‘‘Do Pair Matching’’ button is clicked to automatically assign the subjects into

the study groups (based upon each animal’s response strength) so that group averages and variances are similar prior to treatment. This figure

presents the worksheet, as it would appear after pair matching has been done. See Fig. 2 legend for the details of the study included in this example.

Fig. 5. The daily values of multiple variables measured for each animal are stored in the individual subject worksheets shown here. These worksheets

are filled automatically when the ‘‘Crunch New Data’’ button is clicked. The sheet is organized to contain the results of a single session in each row.

Columns are dedicated for date, group, and session identifiers as well as for each of the measured variables. See Fig. 2 legend for the details of the

study included in this example.


Fig. 6. Whenever the ‘‘Do Group Averages’’ macro updates a group worksheet, each chart in the workbook is automatically updated with the new

values. This allows the researcher to continuously track the course of the experiment as new data points are collected. In this particular example, 24

subjects were injected with either saline or 0.4�/LD50 VX. This graph shows the peak startle amplitude (Newtons) 30 min post-injection for the

duration of the study.

Fig. 7. The statistical output worksheets are formatted for easy import into SigmaStat† or other statistical package. See Fig. 2 legend for the details

of the study included in this example.


Diego Instruments, Inc. Active Avoidance System; the

BioMedic Data Systems’ ELAMS chip system; and the

Med Associates Inc. Operant Conditioning System.

Using VBA macros in conjunction with MicrosoftExcel has allowed us to visualize daily changes in

behavioral data easily, has significantly reduced the

number of data entry errors, has greatly increased the

speed at which the data can be processed (from hours to

seconds), and has freed skilled technicians for more

important tasks. In addition, data files are easily

organized and retrieved, subjects are easily pair matched

and assigned to treatment conditions or test chambers,and data are rapidly transformed for direct import into

statistical packages for rapid evaluation of experimental

outcomes.

Acknowledgements

The United States Army Medical Research and

Materiel Command supported this work under Depart-ment of Defense contract DAAH04-96-0086, Task 685,

administered by Battelle.

Appendix A

Appendix section 1: VBA code to build a workbook

from setup information.A template workbook containing all of the code is

available upon request. In this section, we will present

some of the more interesting code fragments. So that the

code may be discussed in the text, we include a numeric

index with each statement. Although the code fragments

presented in the following sections are numbered con-

secutively, the fragments are not necessarily pulled from

sections of code that are contiguous. Examples oftransferring data to or from the setup worksheet of

the active workbook are given below:

001 NumberOfTreatments�/Worksheets(‘‘Setup Infor-

mation’’).Cells(11,2).value

002 SubjectsPerGroup�/Worksheets(‘‘Setup Informa-

tion’’).Cells(5,2).value

003 FirstSubject�/Worksheets(‘‘Setup Informa-tion’’).Cells(7,2).value

004 NumberOfGroupsPerTreatment�/Work-

sheets(‘‘Setup Information’’).Cells(4,2).value

If the code must specify a cell in a workbook that is

not currently active, then a fully formed cell reference is

required in the format ‘‘Workbooks(wn).Work-

sheets(sn).Cells(r,c)’’ where ‘‘wn’’, ‘‘sn’’, ‘‘r’’, and ‘‘c’’

are variables that contain the workbook and worksheetnames and the desired row and column of the cell. A

reference of this type can be used on the right-hand side

of the equal sign in a VBA assignment statement cause

the value of the cell to be read. Use of the reference on

the left hand side causes a new value to be written to the

cell.

Creating group sheets: The total number of studygroups was defined as the number of groups per

treatment times the number of treatments. The group

sheets were created by copying the ‘‘Group X’’ sheet to a

new worksheet (line 8) and renaming the new worksheet

appropriately (line 9) as shown below:

005 NumberOfGroups�/NumberOfTreatments *

NumberOfGroupsPerTreatment006 For t�/NumberOfTreatments To 1 Step-1

007 For i�/NumberOfGroupsPerTreatment To 1 Step-

1

008 Worksheets(‘‘Group X’’).Copy after:�/Work-

sheets(‘‘Group X’’)

009 Worksheets(‘‘Group X (2)’’).Name�/‘‘Group’’_

& (t�/1) * NumberOfGroupsPerTreatment�/i

010 Worksheets(‘‘Setup Information’’).Cells((t�/1) *NumberOfGroupsPerTreatment_

�/i�/23,7)�/Worksheets(‘‘Setup In-

formation’’).Cells(1�/t,7) & ‘‘�/’’_

& Worksheets(‘‘Setup Information’’).Cells(12�/i,7)

011 Next i

012 Next t

Notice the use of the underscore character on line 10

to indicate that the statement is continued on subse-

quent lines. Since new group sheets were insertedimmediately to the right of the template sheet, the

stepping increment was set to minus one (instead of the

conventional positive one) so that the created sheets

were ordered numerically from left to right in increasing

order. The following fragment was used to clean up the

portions of the groups’ list on the ‘‘Pairmatch’’ work-

sheet and to place a thick border around the groups that

remained:

013 Worksheets(‘‘Pairmatch’’).Range(‘‘I’’ & 2�/Num-

berOfGroupsPerTreatment * _

NumberOfTreatments & ‘‘:O41’’).Delete (xlShift-

Up)

014 Worksheets(‘‘Pairmatch’’).Range(‘‘I2:O’’ & 1�/

NumberOfGroupsPerTreatment * _

NumberOfTreatments).BorderAround weight:�/

xlThick

Creating subject sheets: The number of subjects

required in the study was computed (line 15), and it

was assumed that they were numbered sequentially

starting with the first subject’s number (line 17). The

subject worksheets were then created (lines 18, 19) and

identifying entries were placed into a list on the

‘‘Pairmatch’’ worksheet (lines 23, 24):

015 NumberOfSubjects�/NumberOfGroupsPerTreat-

ment_


* NumberOfTreatments * SubjectsPerGroup

016 For i�/NumberOfSubjects To 1 Step-1

017 subject�/FirstSubject�/i�/1

018 Worksheets(‘‘Subject X’’).Copy after:�/Work-sheets(‘‘Subject X’’)

019 Worksheets(‘‘Subject X (2)’’).Name�/‘‘Subject’’ &

subject

020 Next i

021 For i�/1 To NumberOfSubjects

022 subject�/FirstSubject�/i�/1

023 Worksheets(‘‘Pairmatch’’).Cells(i�/1,1)�/subject

024 Worksheets(‘‘Pairmatch’’).Cells(i�/1,6)�/subject025 Next I

The subject numbers are placed in both columns ‘‘A’’

and ‘‘F’’ of the worksheet because some studies refer to

subjects both numerically and by a name.

Cleaning the ‘‘Pairmatch’’ sheet: The subject list on

the ‘‘Pairmatch’’ worksheet of the template workbook

contained 200 lines of references to subject sheets. The

unused cells from the subject list were deleted (line 27).

Borders were placed around the columns ‘‘A’’ to ‘‘G’’and around column ‘‘H’’ of the remaining section (lines

31, 32):

026 Dim highestitem, lowrow, highrow As Integer

027 Worksheets(‘‘Pairmatch’’).Range(‘‘A’’ & 2�/Num-

berOfSubjects & ‘‘:G200’’).Delete_

(xlShiftUp)

028 For i�/1 To NumberOfGroupsPerTreatment *NumberOfTreatments

029 lowrow�/2�/SubjectsPerGroup * (i�/1)

030 highrow�/1�/SubjectsPerGroup * (i�/1)�/Sub-

jectsPerGroup

031 Worksheets(‘‘Pairmatch’’).Range(‘‘A’’ & lowrow &

‘‘:G’’ & highrow).BorderAround_

weight:�/xlThick

032 Worksheets(‘‘Pairmatch’’).Range(‘‘H’’ & lowrow& ‘‘:H’’ & highrow).BorderAround_

weight:�/xlThick

033 Worksheets(‘‘Pairmatch’’).Cells(i�/1,14)�/lowrow

034 Worksheets(‘‘Pairmatch’’).Cells(i�/1,15)�/high-

row

035 Next i

Assigning subjects to treatment groups: Activation of

the ‘‘Pairmatch’’ worksheet caused it to be displayed,providing feedback for the user during the execution of

the macro. Treatments were assigned to the subject in a

balanced fashion by building a list of sequential integers

into an array called ‘‘treat’’ such that the list contained

equal amounts of each treatment number.

036 Worksheets(‘‘Pairmatch’’).Activate

037 Worksheets(‘‘Pairmatch’’).Cells(1,7)�/‘‘Treat-ment Group’’

038 Dim rndary(3000),treat(3000) As Integer

039 Dim trt, grp, subj, c As Integer

040 c�/0 ‘SET A COUNTER TO ZERO

041 For trt�/1 To NumberOfTreatments

042 For grp�/1 To NumberOfGroupsPerTreatment

043 For subj�/1 To SubjectsPerGroup044 c�/c�/1 ‘INCREMENT THE COUNTER

045 rndary(c)�/c ‘SET AN ARRAY OF IN-

CREASING NUMBERS

046 treat(c)�/trt ‘SET AN ARRAY OF TREAT-

MENT NUMBERS

047 Next subj ‘STEP THROUGH SUBJECTS IN

A GROUP

048 Next grp ‘STEP THROUGH THE GROUPS049 Next trt ‘STEP THROUGH THE TREAT-

MENTS

050 num�/c ‘INITIALIZE A DOWN-COUNTER

051 numsave�/c ‘SAVE THE NUMBER OF AR-

RAY ELEMENTS

The built-in random number generator was initialized

(line 52), and random numbers were generated using the

built-in variable ‘‘Rnd’’ (line 56), which returns a

random value between 0 and 1 each time it is accessed.

Random integers between 0 and the current value of

‘‘num’’ were generated and were used to randomly pick

(line 57) without replacement (lines 58�/61) elements

from the list. The picked value was used to assign a

treatment to the subjects on the ‘‘Pairmatch’’ worksheet:

052 Randomize ‘INITIALIZE

THE RAN-

DOM GEN-

ERATOR

053 c�/0 ‘RESET ‘‘C’’

FOR USE AS

A COUNTER054 For i�/numsave To 1 Step-1 ‘STEP FROM

END TO

START OF

ARRAYS

055 c�/c�/1 ‘BUMP THE

OUTPUT

ROW COUN-

TER056 rval�/Int((num * Rnd)�/1) ‘GET A RAN-

DOM BE-

TWEEN 1

AND ‘‘NUM’’

057 Worksheets(‘‘Pairmatch’’).Cells(1�/

c,7)�/treat(rval)

‘OUTPUT

THAT ELE-

MENT

058 For j�/rval To numsave ‘SLIDE THEREST OF THE

ARRAY FOR-

WARD


059 rndary(j)�/rndary(j�/1) ‘TO PRO-

DUCE A

SHORTER

ARRAYMISSING

060 treat(j)�/treat(j�/1) ‘THE ELE-

MENT THAT

WAS USED

061 Next j

062 num�/num�/1 ‘SHORTEN

THE

LENGTH OFTHE ARRAY

063 Next I ‘AND RE-

PEAT UNTIL

ALL ELE-

MENTS

HAVE BEEN

USED

Creating and customizing the chart sheets: To alter

settings for the charts, a string specifying the data rangein the ‘‘A1’’ format had to be created from column

numbers specified by integers. The following code

fragment provides one way to accomplish the task for

column numbers less than 53. The ‘‘If, Then, Else’’

construct can easily be nested to allow for higher

column numbers. The fragment provides the proper

source string, ‘‘rm’’, that refers to the data from row

‘‘lowitem’’ to row ‘‘highestitem’’ in the column indicatedby the integer ‘‘col’’.

064 Dim lowitem, highitem As Integer

065 Dim sourcespecifier, colstr As String

066 lowitem�/11

067 highestitem�/70

068 If col�/26 Then

069 colstr�/Chr(64�/Int((col�/1)/26)�/(Int((Int((col�/

1)/26)�/1)/26) * 26))_

& Chr(64�/col�/(Int((col�/1)/26) * 26))

070 Else

071 colstr�/Chr(64�/col�/(Int((col�/1)/26) * 26))

072 End If

073 rm�/colstr & lowitem & ‘‘:’’ & colstr & highestitem

The charts were created from a template chart using

copy and rename operations:

074 Charts(‘‘Chart X’’).Copy after:�/Charts(‘‘Chart

X’’)

075 chartname�/‘‘AM 100dB Chart’’

076 Charts(‘‘Chart X (2)’’).Name�/chartname

The chart was accessed using a chart object (line 77)

that was set to a member of the charts collection (line

78) and the chart titles were changed (lines 79�/81):

077 Dim ch As Chart

078 Set ch�/Charts(chartname)

079 ch.ChartTitle.Text�/‘‘title across the top of the

chart window’’

080 ch.Axes(xlValue).AxisTitle.Text�/‘‘y axis title’’

081 ch.Axes(xlCategory).AxisTitle.Text�/‘‘x axis title’’

The individual data series in the chart were accessed

using a series object that was set to a member of the

series collection for the chart (line 84):

082 Dim s As Series

083 g�/4

084 Set s�/ch.SeriesCollection(g) ‘GET THE

4th DATA

SERIES

(GROUP 4)

085 s.Values�/Worksheets(‘‘Group’’ &

g).Range(rm)

‘SET THE

Y-AXISSOURCE

DATA

086 s.XValues�/Worksheets(‘‘Group’’ &

g).Range(rx)

‘SET THE

X-AXIS

SOURCE

DATA

087 s.Name�/Work-

sheets(‘‘Pairmatch’’).Cells(1�/

g,9).value

‘SET THE

SERIESNAME

‘SET THE

ERROR

BAR

SOURCE

DATA

088 s.ErrorBar Direction:�/xlY, In-

clude:�/xlErrorBarIncludeBoth, _Type:�/xlErrorBarTypeCustom,

Amount:�/Worksheets(‘‘Group’’ &

g).Range(re), _

MinusValues:�/Work-

sheets(‘‘Group’’ & g).Range(re)

089 Dim e As ErrorBars

090 Set e�/s.ErrorBars

091 e.EndStyle�/xlCap ‘AND SETTHE ER-

ROR BAR

STYLE

The data range (lines 85, 86), name (line 87), and error

bar values (line 88) were set to values on the ‘‘Group’’

and ‘‘Pairmatch’’ worksheets. The error bar style was set

to the ‘‘cap’’ style (line 90).

Deleting unused groups from the charts: Each of the

ten data series in the template chart points to a similar

plotted variable on the group sheets. If the experiment

did not use all ten groups, some data series needed to be

deleted from the chart. This was done using the‘‘Delete’’ method of the ‘‘SeriesCollection’’ object for

the chart to delete a data series. The following line

deletes the 5th data series:


092 ch.SeriesCollection(5).Delete ‘DELETE AN UN-

USED SERIES IN

THE CHART

Note that after the delete operation the seriesnumbers are recalculated so that what was the sixth

series before line 92 above was executed becomes the

fifth series after execution of the line.

Hiding the template worksheets: To avoid confusion,

the template worksheets were hidden after they were no

longer needed (lines 94�/96). The ‘‘Display Alerts’’

feature of the EXCEL application was toggled disabled

(lines 93, 97) during the operation to avoid a messagebox requesting user confirmation of the actions:

‘HIDE THE CHART, GROUP, AND SUBJECT

TEMPLATE WORKSHEETS

093 Application.DisplayAlerts�/False

094 Charts(‘‘Chart X’’).Visible�/False

095 Worksheets(‘‘Group X’’).Visible�/False

096 Worksheets(‘‘Subject X’’).Visible�/False097 Application.DisplayAlerts�/True

Appendix section 2: VBA program code to find and

read instrument data files.

Only the more interesting aspects of the code are

presented in the interest of space. The complete code

listing is available from the authors.

Improving performance by suspending auto calculation

mode: The EXCEL workbook will evaluate the formula

assigned to a cell and place the new value in the cell

whenever one of the inputs to that formula changes. The

user may not notice the delay when values are entered by

hand. However, when automated entry of data into cells

used as inputs to functions is done, the delay caused by

the repetitive evaluations can accumulate to produce

prolonged and unnecessary delays. The ‘‘autocalcula-tion’’ feature was disabled and the ‘‘calculate on save’’

feature was enabled to avoid the repetitive recalculations

and to ensure that at least one evaluation takes place

before the user saves the workbook.

‘SET EXCEL FOR MANUAL CALCULATION

TO INCREASE SPEED

098 Application.Calculation�/xlCalculationManual099 Application.CalculateBeforeSave�/True

Before subroutine exited, the features were restored.

‘FORCE A WORKBOOK CALCULATION

AND RESET CALCULATION MODE

100 Application.Calculate

101 Application.Calculation�/xlCalculationAutomatic

Finding files to be processed: The method usedassumed that the user had manually moved any raw

data files to be processed into an ‘‘input file’’ directory

specified on the ‘‘Setup Information’’ worksheet. A file

search object (line 102) was used to look for all raw data

files in the input directory (line 103). In this example, the

input files had ‘‘.CSV’’ filename extensions (line 104).

After the file search has been executed (line 105), thenumber of files found (line 106) and the found file names

(line 108) were returned from the object. Any file loaded

by EXCEL becomes the active worksheet so the ‘‘Pair-

match’’ sheet was activated (line 110�/114) after file

processing operations were completed.

‘START LOOP THROUGH SUBJECT’S FILES

FOR USER ENTERED DATE102 Set fs�/Application.FileSearch

103 fs.LookIn�/Worksheets(‘‘Setup Informa-

tion’’).Cells(3,2)

104 fs.Filename�/ ‘‘*.csv’’

‘EXECUTE A FILE SEARCH USING THE

FILESEARCH OBJECT

105 If fs.Execute�/0 Then

‘FOR AS MANY FILES AS WERE FOUND106 filecount�/fs.FoundFiles.count

107 For FileNum�/1 To filecount

108 fname�/fs.FoundFiles(FileNum)

(Process the data file, pulling information from it

and loading the information into the appropriate

rows and columns of the subject sheets. Move the

data file after it has been processed if the user

desires.)109 Next FileNum

‘ACTIVATE THE PAIRMATCH SHEET AND

CONFIRM UPDATES

110 Worksheets(‘‘Pairmatch’’).Activate

111 MsgBox (filecount & ‘‘updated sessions’’)

112 Else

‘ACTIVATE THE PAIRMATCH SHEET AND

CONFIRM NO-FILES113 Worksheets(‘‘Pairmatch’’).Activate

114 MsgBox ‘‘There were no files found.’’

115 End If

Moving raw data files after they are processed: The

example presented in this paper was designed for files

produced by the Hamilton�/Kinder Acoustic Startle

System, which produces comma separated value files

with the raw data. The code to conditionally move

processed data follows:

‘DOES USER WISH TO MOVE INPUT FILES

AFTER THEY ARE PROCESSED?

115 Dim DataMove As Boolean

116 DataMove�/False

117 If Worksheets(‘‘Pairmatch’’).Cells(26,28).value�/

‘‘Yes’’ Then DataMove�/True

After the file was processed:

‘IF USER REQUESTED IT, RENAME THE

CSV FILE TO A NEW LOCATION


118 Dim OldDirectory, NewDirectory

119 OldDirectory�/Worksheets(‘‘Setup Informa-

tion’’).Cells(3,2)

120 NewDirectory�/Worksheets(‘‘Setup Informa-tion’’).Cells(6,2)

121 Dim OldName, NewName

122 OldName�/OldDirectory & ‘‘\’’ & Name

123 NewName�/NewDirectory & ‘‘\’’ & Name

124 If DataMove�/True Then Name OldName As

NewName

Processing each raw data file: The manner in which a

raw data file is processed will depend on many different

constraints. Different types of instrumentation produceraw data files in different formats. Additionally, users

may choose to name data files according to various

criteria or to enter different information into instrument

data fields. Only the most basic processing steps are

described here for the Hamilton�/Kinder (.CSV) data

format.

At our laboratory, the data collection date is encoded

into the file name. The EXCEL program will open commaseparated value files as a workbook with a single

worksheet with the file’s name.

‘OPEN THE CSV FILE (EXCEL

WILL PULL IT INTO A WORK-

BOOK

125 Workbooks.Open (fname)

‘GET THE FILE DATE EN-CODED INTO THE NAME

126 Name�/Application.ActiveWork-

book.Name

127 fileday�/Mid(Name,1,2) ‘DAY

128 fileMonth�/Mid(Name,3,3) ‘MONTH

129 session�/Mid(Name,6,2) ‘RUN NUM-

BER

130 sheetname�/fileday & fileMonth &session

131 indate�/fileday & ‘‘’’ & fileMonth &

‘‘2003’’

‘NOTE THE

HARDCODED

YEAR

‘SORT THE CSV FILE’S WORK-

BOOK BY ENCLOSURE THEN

TRIAL TYPE

132 Selection.Sort Key1:�/Range(‘‘I2’’),Order1:�/xlAscending, Key2:�/

Range(‘‘M2’’)_

, Order2:�/xlAscending, Header:�/

xlGuess, OrderCustom:�/1_

, MatchCase:�/False,

Orientation:�/xlTopToBottom

At this point, the rows in the raw data worksheet were

examined, and the average values for peak response,peak latency and average response were computed. We

ignored the first trial of a particular type and based the

averages on the remaining trials. The averaged values

were saved in arrays and were distributed to the proper

subject sheets after the raw data file was closed:

‘CLOSE THE CURRENT CSV WORKBOOK

WITHOUT SAVING CHANGES133 Workbooks(Name).Close SaveChanges:�/False

Distributing data to the subject sheets: This relatively

straightforward process requires no example code frag-

ments. The only point of interest is how multiple

sessions in a single day were handled. Each animal in

the example study (see Fig. 2) was tested twice a day

(once in the morning and once in the afternoon). The

startle system used during the study had eight separate

enclosures so that eight subjects could be tested in asingle session. Running 16 sessions per day allowed a

maximum of 64 animals to be tested twice a day (eight

morning sessions and eight afternoon sessions). The 16

raw data files produced each day were assigned session

numbers encoded into the file name. Data from the first

eight sessions were stored in the ‘‘AM’’ data columns on

the subject sheets while the data from the second eight

sessions were stored in the ‘‘PM’’ data columns. The

column set to be used was computed from the session

number. Each row on a subject sheet corresponded todata collected on a particular date, which was stored in

the first column of the row. Given the date that the data

was collected (from the raw data file name), the program

simply scanned down the first column looking for a date

that matched the test date in order to locate the row into

which the new data should be placed. If no matching

row was found, the next available empty row was started

for the test date.

Processing data files that EXCEL cannot open directly:

Some instruments produce data files that cannot be

opened conveniently by EXCEL. We have included a

fragment of VBA code that can be used to open text

files.

‘TO OPEN A FILE

NAMED fname

134 Open (fname) For Input

As #1‘TO READ THE NEXT

LINE FROM FILE (TO

LINE FEED) INTO A

STRING

135 Dim InputLine as String

136 Line Input #1, InputLine ‘THIS STATEMENT

WOULD BE RE-

PEATED‘TO GET DATE FROM

SOME PART OF THE

LINE JUST READ

137 Dim part as String


138 part�/Mid(Input-

Line,10,8)

‘GET 8 CHARAC-

TER STARTING AT

POSITION 10

‘TO CLOSE THE FILEAFTER IT HAS BEEN

READ

139 Close #1

Appendix section 3: VBA program to perform group

averages.

This straightforward procedure averages the dataitems from corresponding dates and columns for the

subjects that are members of a group and places the

average, count, and standard deviation of those values

into the group sheet for that group. The process is

repeated for each date of each group. No code fragment

is given. It is critical to disable the auto calculation

feature beforehand and to restore it afterwards. Other-

wise, a complete set of chart updates (based on new data

in the group sheet) will be done for each entry onto the

group sheets. The only interesting point is that the user

must ensure that appropriate subject rows are included

in the averages. In the case described in this paper, all

subjects were treated on the same daily schedule. If there

is more than one daily schedule in a study, we have

found it useful to define a ‘‘study day’’ column that can

be used with a listing of subject starting dates to preserve

the study’s time information. In such a case, the group

average should include same study day across the

group’s subjects.

References

Jacobson R. EXCEL 2000 VISUAL BASIC FOR APPLICATIONS: Funda-

mentals. Redmond, WA: Microsoft Press, 1999.

Wells E, Harshbarger S. Microsoft† Excel 97 Developer’s Handbook.

Redmond, WA: Microsoft Press, 1997.


behavioral data management using visual basic for applications to automate data capture and analysis

Documents