Survey Development

Our experts are dedicated to collecting information from people about their ideas, feelings, health, plans, beliefs, and social, educational, and financial background. It usually takes the form of self-administered questionnaires and interviews.

All surveys rely on directly asking people questions to get information. Survey must be concerned with who and how many will be asked questions (sampling), when and how often (design), and processing, analysis, and interpretation of data. Survey purposes and methods fall on a continuum. Some surveys can have far-reaching, generalizable effects, and their methods must be scientific.

All surveys consist of (1) questions and (2) instructions, and they make sense only in the context of (3) sampling and design, (4) data processing and analysis, (5) pilot testing, (6) response rate, and (7) reporting results.

The Survey Form

We are experts in defining the attitude, belief, or idea being measured. Survey questions may be forced choice or open ended. Forced-choice questions with several choices are easier to score than are open-ended, short-answer, essay questions. Open-ended questions give respondents an opportunity to state a position in their own words.

When writing forced-choice or open-ended questions, we use standard English, keep questions concrete and close to the respondents’ experience, become aware of words, names, and views that might automatically bias our results, check our own biases, and use a single thought in each question.

The responses to closed-ended questions can take the form of yes-no answers, checklists, rating scales, comparative rating scales, and category scales. Rating scales may be graphic, but often they ask respondents to make comparisons in the form of ranks (1 = top to 10 = bottom), orders (1 = definitely agree, 2 = agree, 3 = disagree, 4 = definitely disagree), or continuums (your age). The numerical values assigned to rating scales can be classified as nominal, ordinal, interval, or ratio (numerical).

Surveyors are most often interested in responses to individual items such as the number of people who will vote for Candidate X. Making scores meaningful by proving that high scorers are truly different from low scorers requires knowledge of (additive) scaling methods. Three commonly used methods create differential (Thurstone), summated (Likert), and cumulative (Guttman) scales.      

Sampling

Sampling can be divided into two categories. A probability sample is selected by an objective method, and we can also calculate each person's chances of selection. Nonprobability samples are convenient: We select only those respondents who are willing and available to complete the survey. Nonprobabiltiy samples are usually easier to assemble than probability samples. But gains in ease can be met with loses in generalizability.

Probability Sampling Methods

Nonprobability Sampling Methods

We can ask five questions when determining sample size:

What is the null hypothesis?

What is the desired level of significance related to the null hypothesis involving the mean in the population?

What differences between means are important?

What is a good estimate of the standard deviation in the population? 

Survey Design

A survey’s design describes the frequency of administration (twice), when the administration takes place (before and after), and the number of groups to be surveyed.

A cross-sectional design provides a portrait of things as they are at a single point in time. A poll of voters’ preferences one month before an election and a survey of the income and age of voters in the same election both use cross-sectional designs. Longitudinal surveys are used to find out about change. Trend designs are longitudinal. Another longitudinal design is the cohort. A third longitudinal design is the panel.

In comparison group survey designs, the groups we survey can be assembled randomly or in some other way, perhaps voluntarily. Random assignment usually makes it easier to draw valid conclusions from survey data. Normative designs take two forms. In the first, two groups are compared, but only one is actually surveyed; the other group, the comparison, consists of data that are already on record. The second type of normative design uses a "model” as a standard for comparison.

A case control design is one in which groups of individuals are chosen because they have (the case) or do not have (the control) the condition being studied, and the groups are compared with respect to existing or past attitudes, habits, beliefs, or demographic factors that are judged to be of relevance to the causes of the condition.

Data Analysis

We can design and perform the required statistical analyses. We are experts in statistical analysis such as t-test, ANOVA, traditional regression, logistic regression, MANCOVA, factor analysis, cluster analysis, survival analysis, time series analysis, canonical correlations, discriminant analysis, and more advanced statistical techniques such as structural equation modeling (SEM), hierarchical linear modeling (HLM), and path analysis. We are experts in statistical programming languages such as SAS, SPSS, HLM, Mplus, and Access.

Presenting the Survey Results

Survey results can be shown on the survey form itself and in tables, graphs, diagrams, or pictures. This approach shows the survey’s raw data. Survey reports nearly always have tables. Their purpose is to describe respondents and show relationship and changes. We select a table format and use it consistently, rely on vertical lines sparingly, and present data in logical order such as from most to least frequent. Most surveys result in written reports. We consider including these sections in our report: abstract, summary, table of contents and figures, glossary of terms, statement of purposes or objectives, methods, results, conclusions, recommendations, references, index, appendix, and acknowledgments. Overheads are effective visual aids in oral survey reports.  

Confidentiality

We provide all of our prospective clients with a Non-Disclosure Agreement immediately upon contact. Statistical Consulting Group shall maintain in strict confidence, and not use or disclose except pursuant to written instructions from prospective clients, provided that the obligation to protect the confidentiality of any such information or data shall not be excused if such information or data ceases to qualify as a Trade Secret as a result of the acts or omissions of Statistical Consulting Group.

Statistical Consulting Group
1050 WINTER STREET SUITE 1000
WALTHAM, MA 02451
PHONE: 781-205-4719
FAX: 781-780-7918
info@statisticalconsultinggroup.com