|
Survey Analysis |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Introduction
|
Surveys are widely used by commerce and industry to measure a range of responses, including community, customer and staff attitudes, customer service satisfaction, product acceptance and corporate culture.
They can be expensive, prompting some organisations to consider in-house administration and analysis. This often raises issues of confidentiality, confidence in the data, the design of the survey and interpreting the analyses. Irrespective of the way the data is collected and how well the survey is conducted, the system described in the following can assist in overcoming many of the technical issues.
The Gryphon Survey Analysis Database is designed to store and analyse data captured from a variety of attitude surveys, questionnaires and tele-marketing exercises.
The system eliminates the need to construct a database or statistical analysis file, re-code data or write new procedures each time a survey is undertaken. The intelligence built into the software aids the designer in developing the types of questions that may be used in the survey, while the battery of statistical, descriptive and graphical procedures makes it simple to produce high quality reports.
The advantages of the system are;
the same database can be used for just about any survey it can handle all common survey data types there is no need for users to understand database technologies it can be operational in 10 minutes it allows a range of data capture methods from direct entry to web based capture it is simple to use yet has a wide range of analysis procedures available it allows for the storage and analysis of time series data the product and the 4GL in which it is written is Australian owned and developed. The following points could be considered:
1. Set down the precise objectives of the survey 2. Determine who is to be surveyed. If a sample is to be taken determine the required sample size, method of selecting the sample, and the population from which the sample is to be drawn. 3. Determine the method of administering the survey (interview, questionnaire, combination, observation) 4. Determine the treatment of errors (sampling, inaccuracies, bias, response) 5. Determine in general terms the analysis to be undertaken. This phase may well give rise to new objectives or a rethink of what data is being collected. 6. Develop the questions. At this point you are ready to create the database. 7. Pilot the survey to test assumptions, variability of responses, cost and time in administering, etc.
The SAD is suitable for analysing all forms of questionnaires whether they be complex market research exercises or simple information gathering such as capturing customer demographic details. The system provides for structured and unstructured questions, along with 'agree/disagree' type questions which attempt to gauge the strength of attitudes. It can therefore be used to store and analyse general text responses, specific answers and coded answers including multiple choice and ranked items.
Telemarketing is a specialised form of questionnaire conducted over the telephone, usually from a script with responses entered directly into a data storage facility. The SAD can be used in this mode.
Social surveys commonly take the form of a census or community study, often using sampling techniques. The SAD can accommodate the usual social survey instruments as well as provide time series or trend analysis where the survey is administered more than once.
The database will be initialised and the operator placed in the question format screen prompted for the question type of the first question. Press {ESC} to exit if you are not ready to enter question types.
Time series data analysis can be undertaken and each survey is held as a discreet entity indexed by a date. Follow-up surveys can be identical with a new date as the index.
Question Type
There are three phases involved in data entry, two of which are mandatory For each question or part-question, specify a question type, a question label and valid range for appropriate question types. It is important that even where the response is an alpha entry, such as the code AAAA, a range defining the number of different valid codes should be entered. The grouping or questions into analysis dimensions is allowed at the completion of question loading. This phase is best undertaken at the time of designing the survey as it forces the designer to consider the types of reporting to be conducted on the captured data. Default ranges are provided to those questions requiring specified ranges. These may be overwritten by the operator. The labels entered are limited to 50 characters in length and are converted to uppercase for use in graphs, tables, etc.
The types of questions supported are: Likert style comparative on 5 to 9 point scales - between 'what should be' and 'what is' 1 Yes/No or True/False answers - Three responses are accepted - yes (1), no (0) or don't know (2). 2 Likert style answers - on a 5 to 9 point scales of the form Disagree through Agree 3 Numeric Values between 0 and 999999 4 Dollar Values between 0.00 and 9999.99 5 Two digit coded numeric values between 0 and 99 6 Multiple Choice up to 50 sub-variables chosen by entering a 0 for no response or 1 if ticked. 7 Coded Alpha will accept any number of characters up to 50. For processing considerations, it is desirable, to limit the number of characters entered. A range must be entered indicating the number of valid codes. 8 Ranked choice type question accepting a ranked list up to 50 items in length. Again a range needs to be entered indicating the number of items. 9 Free text up to 999 lines. Enter answers for each respondent against each question. This is where the response is entered against each respondent identification number. In simple surveys requiring yes/no or Likert style answers, the operator simply types the Respondent Number, Question Number or hits {CR} followed by the answer. The default operation increments the question numbers. Press {ESC} to enter the next respondent or press {ESC} again to exit back to the Menu.
Nine analysis categories are provided. Some analysis forms are inappropriate for certain types of question or data as depicted below: For example, it is clearly not possible to cross tabulate a continuous variable (types 3 and 4 that may have any value between 0 and 999999). Crosstabs rely on the variables having discrete fixed values which are summed for each cell. The reports category produces straightforward listings of the data whether it be free text entered against the type 9 questions or a score of 1 to 5 for type 2 questions. The ranking report is another which simply ranks in order the re-coded responses.
The QWE (Quality of Work Experience) report is a special case which makes use of the attitude dimensions specified and the type 2 questions. The responses are re-coded to a positive or negative score of up to 100 then averaged for each dimension. The scores provide an indication of the relative strength of the dimensions.
The forms of trend analysis are variations of the specific analysis programs.
The primary goal of the descriptive statistics provided in most of the analyses is to bring order out of chaos. The SAD provides a means of describing results using regular and cumulative frequency distributions, measures of central tendency (mean, mode, median), measures of variability (range, standard deviation, variance) and percentile scores. Inferential statistic techniques for drawing conclusions about an entire population based on a sample are also available in the form of linear regressions (correlations). ^
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
