Frequently Asked Analysis Questions

1. How can I change the appearance of the printouts?
2. How can I do an analysis on a subgroup of the data (e.g., just females)?
3. How can I do an analysis on more than one subgroup?
4. How can I create groups (categories) for interval data?
5. How do I keep track of all the options?
6. How do I deal with the open-ended questions?
7. How can I get the Input fields in the Data Manager to line up properly?
8. How can I get the Data Manager to accept data for a particular variable?
9. How can I do the same study again with a new set of data?
10. How can I create a new study that is similar (not identical) to a previous study?
11. Can I change the codebook and form after entering some data?
12. How can I import a fixed format sequential ASCII data file into StatPac?
13. How can I fix a "duplicate variable" error message?
14. How do I combine data files so they can be analyzed together?
15. How do I deal with an "other" category?
16. I have more general questions about StatPac for Windows.
17. I have questions about Web surveys.

QUESTION 1 - My printouts are close to what I want, but not exactly. What can I do to change the way StatPac works?

ANSWER - Options are used to control the way that StatPac formats reports. While using the analysis procedure file editor, move the cursor to the procedure you want to modify. Select Options to evoke the Options window. Change the desired option(s) and click OK.

QUESTION 2 - I want to perform an analysis using just the female respondents. What line do I include in the procedure?

ANSWER - To perform an analysis on a subset of data (e.g., just the females), use the SELECT command in the same procedure as the analysis command. For example, if Sex had been coded as M=Male and F=Female, the command might be:

IF Sex="F" THEN SELECT
FREQ Some_Other_Variable
..

Quotation marks are used around the "F" because Sex was coded as an alpha variable. Numeric variables do not need quotation marks around the selection criteria. For example, if Sex had been coded as 1=Male and 2=Female, the command would be:

IF SEX=2 THEN SELECT
FREQ Some_Other_Variable
..

QUESTION 3 - I want to perform an analysis on respondents that make at least $20,000 per year. What line do I include in the procedure?

ANSWER - This is essentially the same as the previous problem. The difference being that we're selecting multiple categories, instead of just one. Suppose the Income variable is coded as:

1=Under 10,000
2=10,000 - 19,999
3=20,000 - 29,999
4=30,000 - 39,999
5=40,000 or more.

To select respondents making 20,000 or more, we need to select response codes 3, 4 and 5. Any of the following lines could be used to solve the problem:

IF Income>2 THEN SELECT
IF Income>=3 THEN SELECT
IF Income>=3 AND Income<=5 THEN SELECT
IF Income=3 OR Income=4 OR Income=5 THEN SELECT
IF Income="3/4/5" THEN SELECT

When using the slash to indicate a series of OR statements (last example), quotation marks are required for both alpha and numeric-type variables.

QUESTION 4 - I want to use an Age variable as one of the points in a banner table, but age was entered as the actual age instead of being coded into categories. How can I make it work in a banner table?

ANSWER - There are many different solutions to this problem. In order to do this, you need to define age groups. The first solution is to recode the Age variable, and then assign value labels to the recoded data. Two lines are required.

RECODE Age (LO-20=1)(21-30=2)(31-40=3)(41-HI=4)
LABELS Age (1=Under 21)(2=21-30)(3=31-40)(4=Over 40)

The recoded and newly labeled Age variable could then be used in a banners or crosstabs table.

If you wanted access to both the raw data and the recoded data, you could create a new variable and have it contain the recoded age data. This technique has the added advantage that the original data remains intact even when the SAVE command is used. Here are three more solutions to this problem:

LET Age-Group=Age
RECODE Age-Group (LO-20=1)(21-30=2)(31-40=3)(41-HI=4)
LABELS Age-Group (1=Under 21)(2=21-30)(3=31-40)(4=Over 40)
..

COMPUTE (N2) Age-Group=Age
RECODE Age-Group (LO-20=1)(21-30=2)(31-40=3)(41-HI=4)
LABELS Age-Group (1=Under 21)(2=21-30)(3=31-40)(4=Over 40)
...

NEW (N1) "Age-Group" Age Group
IF Age<21 THEN COMPUTE Age-Group=1
IF Age>=21 AND Age<=30 THEN COMPUTE Age-Group=2
IF Age>=31 AND Age<=40 THEN COMPUTE Age-Group="3"
IF Age>=41 THEN COMPUTE Age-Group=4
LABELS Age-Group (1=Under 21)(2=21-30)(3=31-40)(4=Over 40)
..

QUESTION 5 - There seem to be so many options for some of the analyses. How do I keep them all straight?

ANSWER - This one's easy...you don't have to. There are only a few options that have profound impact on the way analyses are performed. The default values for the options are sufficient for most applications. When in doubt, first try running the procedure and make note of what you would like to change. Then select Options to see if the appropriate option is readily apparent. If not, use the on-line help to review the analysis. After you've found the option you are looking for, add it to the procedure to confirm that it does what you want. Finally, if this is something you want to change on all future analyses, add an exclamation point suffix to the option and rerun the procedure. This will make the current option setting the default and you won't need to be concerned with that option again. After running a few procedures, you'll have configured the default formats for StatPac to produce the reports you most often use.

QUESTION 6 - How do I deal with the open-ended questions?

ANSWER - You can do two things with open-ended data. The first is to just list the actual comments. Use the LIST keyword to show the actual verbatim text. If the variable were called Comments, the command would be:.

LIST Comments
..

The output from the procedure would print dashes for respondents who made no comment. The IF-THEN command could be used to limit the output to only those respondents who made a comment. These two procedures would produce identical output. The first one selects all records where the Comments variable is not blank. The second one rejects all records that are blank.

IF Comments <> " " THEN SELECT
LIST Comments
..

IF Comments = " " THEN REJECT
LIST Comments
..

Another solution is to code the responses into categories. Run FREQUENCIES on the open-ended comment and set the OE option equal to Y. This will evoke StatPac's coding program (Verbatim Blaster). Your procedure might look like this:

FR Comments
OPTIONS OE=Y

StatPac's Verbatim Blaster module is easy to use, but will be even easier if you first read about it in the manual. You'll find complete information on-line under FREQUENCIES with the OE option.

QUESTION 7 - I'm using the Data Manager and the program is acting goofy (things are not lining up properly). What did I do wrong?

ANSWER - The most likely problem is that the data entry form was compiled on a different computer with a different screen resolution. Close StatPac. Delete the "codebookname.frc" file. Then run StatPac again and the problem will correct itself.

QUESTION 8 - I'm using the Data Manager and one of the data input fields won't accept data even though the data is correct. How do I fix the problem?

ANSWER - The codebook is somehow limiting the data input for that variable. Examine the variable with the Variable Detail window or the Grid. Check the valid codes. A common problem is where the valid codes are lower case, but the data entry control parameters have Caps Lock set. No matter what the data entry person types, it is converted to upper case, and the only valid codes are defined as lower case. To fix the problem, either change the valid codes to upper case, or set Caps Lock off. Another common problem is having defined a field as numeric, when the data actually contains numbers and letters. For example, many internal ID codes use numbers and letters. The solution is to change the format for the variable from "N" to "A". Another common problem is to have defined some of the valid codes (but not all of them).

QUESTION 9 - I have finished analyzing a study and I want to do the same study again with a new set of data. How can I do this without retyping the study design?

ANSWER - If the new study uses exactly the same variables, then the easiest way to do this is to rename the existing data file to something else. For example, if you conducted a survey called "Opinions" in 2002 and you wanted to do the same survey in 2003, you might rename the 2002 data from "Opinions.dat" to "Opinions-2002.dat" and then begin entering the new data. To rename the old data file, select File, Open, Data File, and right click on the file to be renamed.

If for some reason you wanted to analyze the old data file, use the DATA command to specify the old data file (with a different name than the codebook).

STUDY Opinions
DATA Opinions-2002

QUESTION 10 - I need to do a new study similar to a previous study, but some of the questions are different. Do I have to retype all the study information?

ANSWER - No. There are two ways to do this. The first way is to first create a duplicate copy of the previous codebook using a new name. Load the codebook and then select File, Save Codebook, and give it a new name. You can then modify the new codebook file. The other way to do this is in the Grid. Load the previous codebook as a library (select File, Open Library) and you will be able to extract selected variables from the previous codebook into the Grid.

QUESTION 11 - I designed a codebook and form and tried entering a few data records and discovered that I need to change something in the study design. How can I do this after the data file already has data?

ANSWER - Once data has been entered into a data file, you can still change any text in the codebook or data manager form. You can also change the field width of any variable. If you change the field width of a variable, the associated data file will also be adjusted when you save the codebook..

However, once a data file exists, you should not 1) add a new variable, 2) delete a variable, or 3) change the order of variables, because the data file would no longer match the revised codebook. When you load a codebook that has an associated data file, StatPac will give you a warning that lets you turn off the buttons for these operations. (You can turn off the warning by selecting Format, Codebook Safety, No).

If you don't need the data file (e.g., it's just dummy test data), simply delete the data file and then make the desired changes to the codebook You will also have to modify the data entry form because it does not automatically reflect the changes in the codebook. The easiest way to change the form is to simply delete the existing form and recreate it after the codebook has been modified. To delete the form, select Edit, Select All, and then click the Cut button or select Edit, Cut. Make the changes to the codebook with the Variable Detail window or the Grid and then recreate the form.

If you have already entered a substantial number of real data records, and then discover you need to add a new variable, you must run an analysis to create the new variable in both the study design and data file. The form however, will not automatically be updated to reflect the new variable. Therefore, you must either delete the form (see above) and recreate it, or you can manually insert the new variable in the proper place of the form. A new variable can be created by running a three-line procedure:

STUDY CodebookName
NEW (Format) "New_Variable_Name" New Variable Label
SAVE
..

For example, in a study called "Research", you could create a new numeric two-column variable called Number, with the following commands:

STUDY Research
NEW (N2) "Number" Special New Variable
SAVE
..

When you run the procedure, both the codebook and data will be updated to include the new Number variable. The new Number variable would be added to the end of the existing codebook. You can insert the new variable in the middle of the codebook by using the WRITE command instead of the SAVE command. If you had 100 variables in the codebook and you wanted the new Number variable to become the 26th variable, the commands would be:

STUDY Research
NEW (N2) "Number" Special New Variable
WRITE Research V1 - V25 Number V26 - V100
..

The above commands are "risky" because a mistake in the WRITE command might cause a loss of data. For example, if you made a typo in the WRITE command and inadvertently left out the number "5", there would be a loss of variables 3 through 25 in both the codebook and data files:

WRITE Research V1 - V2 Number V26 - V100 (Bad)

A safer way is to use the WRITE command to create a new codebook and data file rather than replacing the existing file. The following procedure creates a new codebook and data file called RESEARCH-2. The orignal codebook and data file will not be altered. You could then create a data entry form for RESEARCH-2:

STUDY Research
NEW (N2) "Number" Special New Variable
WRITE Research-2 V1 - V25 Number V26 - V100
..

For more information on creating and saving new variables, see the NEW, LET, COMPUTE, and WRITE commands.

QUESTION 12 - How can I import a fixed format sequential ASCII data file into StatPac?

ANSWER - Data files already in sequential ASCII fixed format do not usually need any conversion to be used by StatPac. The only requirement is that they have a .dat extension. First, design a codebook with StatPac so that the field widths for each variable are exactly the same as the data file you will be using. Then copy the data file to the same folder as the codebook and rename it so it has the same prefix as the codebook and a .dat extension. That's all there is to it.

As a final check, you can run the utility program to check a codebook and data file for errors (Analysis, Utilities, Codebook, Check Codebook and Data). If no errors are reported, the length of each data record matches the length defined by the codebook, and you can be assured that the codebook design and data file will interface properly.

If a data file is in any format other than fixed format sequential ASCII, you must use the import program to import the file. The import program will create a new data file and new codebook that best accommodates the format of the data being imported. You could then add variable and value labeling to the codebook.

QUESTION 13 - I ran an analysis that creates new variables and found that some of the computations were wrong. I fixed the procedure and tried to re-run it but it keeps giving me a "duplicate variable" message.

ANSWER -When you used the SAVE keyword, the new variables were saved in the codebook and data file...even though the computations were not correct. When you try to re-run the procedure, StatPac thinks you are attempting to create more new variables with the same names. To fix the problem, comment out the NEW keywords by placing an apostrophe at the beginning of the NEW lines. Then re-run the procedure.

QUESTION 14 - I have data files that I want to merge together. How can I combine them into one file so that I can analyze them together as one group of data?

ANSWER - When the data files contain identical variables in the same order, the data files can by joined by a process called concatenation. Usually, the data files are another administration of the same survey to a different group of people. To merge the files, in the Analysis program, select Utilities, Merge, Concatenate Data Files. For example, suppose you administered a consumer survey at three different shopping malls. The data file names are Mall-1.dat, Mall-2.dat, and Mall-3.dat. The codebook was the same for all three surveys (Mall-Survey.cod).

In many studies, it may also be important to know which data file each record came from after the data is merged. For example, if the purpose of the study were to identify differences between the shoppers at the different malls, it would be important that the final merged data file contained a variable that identified each respondent from Mall 1, 2 or 3. If the original data files do not already contain a group identifier variable, it can be added by running procedures to create the new variable, assign value labels to it, and assign the appropriate value to each record in the data files. This must be done before concatenating the data files. For this example, the procedures might be:

STUDY Mall-Survey
DATA Mall-1
COMPUTE (N1) Location = 1
LABELS Location (1=Mall 1)(2=Mall 2)(3=Mall 3)
WRITE File-1
..
STUDY Mall-Survey
DATA Mall-2
COMPUTE (N1) Location = 2
LABELS Location (1=Mall 1)(2=Mall 2)(3=Mall 3)
WRITE File-2
..
STUDY Mall-Survey
DATA Mall-3
COMPUTE (N1) Location = 3
LABELS Location (1=Mall 1)(2=Mall 2)(3=Mall 3)
WRITE File-3
..

After running the three procedures, each data file would contain a new variable called Location. This variable would have a value of 1 for all the records in the File-1.dat, 2 for all the records in the File-2.dat, and 3 for all the records in the File-3.dat.. The data files could then be concatenated into a single large data file. You would concatenate File-1.dat, File-2.dat and File-3.dat into a new data file called Combined.dat. Subsequent analyses could use the DATA command to specify the concatenated data file as the one to be analyzed. All three codebooks (File-1.cod, File-2.cod and File-3.cod) would be identical, so you could use any of them in the STUDY command.

STUDY File-1
DATA Combined

The concatenation method of creating a merged data file is used only when the data files contain exactly the same variables in the same order.

Often, subsequent administrations of a survey will contain revised and new questions. Although most of the information is the same, small differences make it impossible to just use concatenation The first step in merging this kind of data is to identify the variables that are common to both surveys. Next, run one or more procedures using the WRITE keyword to create subfiles that only consist of the common variables (those variables that are identical in each study).

STUDY FirstStudy
WRITE File-1 V4 V2 V12 - V22
..
STUDY SecondStudy
WRITE File-2 V7 V2 V15-V25
..

Finally, use concatenation to merge the data files. In this example, you would merge File1.dat and File-2.dat into a new larger file called CommonVars.dat. You could then perform the analysis on the combined data by beginning the procedure file with these commands:

STUDY File-1
DATA CommonVars

The other type of merge that StatPac can perform is for a matched pairs or pre/post type of experiment. When two or more data files represent the same individuals, but different variables, you should not use concatenation Instead, you would use the MERGE keyword. Examples are before and after surveys, client follow-up surveys, etc. The data files may contain the same or different variables. The key point is that information on a particular individual is in each data file.

If each data file contains exactly the same number of records in the same order, you can use the MERGE command to merge the files without any additional steps. For example, suppose you have pretest and posttest surveys for the same group of people and the records in the data files are in the same order (i.e., the first record in the PreTest file is John Jones, and the first record in the PostTest file is John Jones; the second record in the PreTest file is Mary Smith, and the second record in the PostTest file is Mary Smith, etc.). The command to merge the posttest data into the pretest data would be:

STUDY PreTest
MERGE PostTest
..

If the data files contain records that are not in the same order, or if there are not the same number of records in each data file, then all the files must contain a unique ID variable that will allow you to match up data in the files. First you will have to sort all the files by the ID variable. It is OK if a data file does not contain a matching record for each record in the other file(s) (e.g., a respondent dropped out of the experiment). They just have to be in ascending sorted order. An example would be:

STUDY PreTest
SORT (A) ID
SAVE
..
STUDY PostTest
SORT (A) ID
SAVE
..

After the data has been sorted, select Analysis, Utility, Merge, Restructure/Merge to perform the merge using ID as the common variable. That is, the ID number will be used to match up the records from each data file. For more information on the merge utility program, see the Utilities section of the on-line help.

QUESTION 15 - One of the variables in my study has an "other" category and a place for the respondent to write in their answer. How do I include this is the study design?

ANSWER - A variable in StatPac is a "piece of information". In this question, you really have two pieces of information, and thus, two variables. The first variable is the one with the defined response codes. The second is the "other" response.

For example, the following question has four response categories:

Who would you vote for to be the next president of our club?

1=Sally Nelson
2=Jim French
3=Brenda Johnson
4=Other Specify: __________________________________

Two variables would be specified in the study design to hold the information for this question. The first variable would be a numeric one-column variable, and the second would be an alpha variable (about forty columns). Skip codes could be used to bypass the open-ended variable if 1, 2, or 3 is entered for the first variable.

1=Sally Nelson ;3
2=Jim French ;3
3=Brenda Johnson ;3
4=Other Specify: __________________________________

Copyright Š 2013 StatPac Inc., All Rights Reserved