An introduction to analyzing survey results
When we receive and collect enough responses for a survey, it is time to analyze the data. Actually, being able to analyze these data in a proper way, is even more important than designing the questionnaire itself.
There are many different ways through which one can extract a proper analysis from the collected survey data. All these ways will lead to achieving a realistic insight about the responses. In this article we are going to have a quick and elementary review of the essential steps and techniques needed to analyze the survey results.
In this article:
Data validation | Segmentation | Coding | Statistical inference | Descriptive analysis | Data types | Conclusion
When a quantity of responses that we intended is gathered, we need to “purify” them. Actually it is required, to have a complete and consistent dataset in order to start working with and do the analysis on.
Therefore as a first step, you can exclude the incomplete data (results from surveys which are not answered by some respondents) from your analysis, because they may cause bias in your results.
The reason for this potential bias is that this incomplete data probably is not an accurate representation of the investigated population. Thus considering it in our calculations may cause bias in the analysis. As an illustration, suppose that for a certain question we have only 50 responses out of 1000 respondents (i.e. 5% of our underlying samples). It is clear that this 5% is not a reliable proportion, to consider it as the representation of our entire population.
One of the a few exceptions that you can include incomplete surveys in your final data analysis is when you are sure you have collected enough information from a participant, that is crucial for your analysis. As we explain here in this article about “creating professional surveys”, it’s strongly recommended to put questions with less importance at the end of the survey, and important questions close to the beginning. This is to make sure you collect the vital data needed for your analysis, even if respondents don’t completely finish the survey, or don’t even submit it.
Sometimes partitioning the data based on demographic categories (like gender and age) may help you boost your analysis and obtain better results.
The next step is to manipulate and convert your data. Different type of the data requires different type of analysis. Unlike numerical scale data, we need to convert nominal and ordinal data in such a way that can be handled using our underlying statistical package. We will discuss this converting techniques later.
Additionally, it would be nice to generate some categories for your data. It is usually desirable to focus on different parts of our survey responses separately. Categorizing the responses gives us this opportunity to analyze our data in each segment.
For instance, comparing the responses for a certain category leads us to a better understanding of our customer’s opinion and therefore helps the management team to make a better decisions about their strategies. Examples of such kinds of categories in a customer satisfaction survey could be the “customer service category”; or in “market segmentation” the common categories are gender and age. Coding provides an aggregation of data into much small and therefore more useful categories.
Categories should be meaningful for your work and should be chosen carefully. You should always think about the reason you are defining a category, why you need it and what you want to do with that information.
A very desirable place to start our analysis is using the descriptive statistics i.e. summarizing the samples and determining the main features of the data e.g. percentage, mean, median etc. The good point about the descriptive analysis is that it return a simple overview about our underlying data and can be visualized easily.
Application of descriptive analysis depends on type of the survey method and response formats. As you know, there are many different question types that you can utilize in an online survey. Usually, it is possible to visualize collected data for all of them, with help of diagrams and charts. To help you get a nice and clear Descriptive Analysis, SurveyLegend offers lots of different data visualization possibilities. Just open your Live Analytics, and enjoy previewing your data in different formats.
The next step in our analysis is to go a little bit deeper in the data and try to understand what do responses really mean. Actually, descriptive analysis alone, cannot illustrate all facts about our data . Therefore, some more advanced methods need to be applied in order to get a better understanding.
You can for example use a hypothesis test, to check if the responses between two of your groups are statistically equivalent or not; or you can use ANOVA to check multiple groups at the same time. One may use a data mining method like regression in order to generate a predictive model as well.
All in all, statistical inference is an undeniably important part of data analysis that is needed to be applied carefully and professionally. You can always get benefit of consulting with an expert for doing this type of analysis.
So, survey data is eventually all about the statistic. Most of the people think that it is enough to look at the descriptive analysis, and extract the knowledge from the data; which is not always true! Actually, we throw away a lot of information if we don’t use statistical inference.
In this section we have just touched the surface of this concept and our goal is to show you as the reader, that there is much more you can do with the underlying data.
There are different datatypes that a survey may include, and each of these types has its own properties which need different type of methods to group and analyze. Therefore, it is important to determine which kind of datatype we are dealing with when we start our analysis. Each one of the questions and responses in a survey can be categorized in one of the following types of data:
- Nominal (categorical) data e.g. gender, marital status,
- Ordinal data e.g. ranking an opinion between 1 to 10
- Interval data e.g. temperature, age,
- Numerical (ratio) data which are just real numbers
After determining the data type, we will be able to analyze the data in a good way.
Please click on the data types, to read more about them in details.
Nominal data Ordinal data Numerical (ratio) data Interval data
Online surveys are effective tools for gathering information from the target population (for example customers) and knowing their opinions. There is no doubt about the importance of having a well-designed questionnaire and adopting an appropriate sampling method. However it is also vital to analyze the responses in a proper and professional way.
Different types of data should be treated differently and you should always consider the goal of your analysis when you implement the statistical tools. Sometimes segmentation and grouping the respondents gives you a better insight about the data and boost your analysis.
When you finish with the preliminary data interpreting e.g. generating contingency tables and doing the descriptive statistics, it is the time to implement some more advanced statistical tools, namely hypothesis test and regression, to obtain a better insight about your data.
Do not forget that the purpose of surveying is to achieve some new information and knowledge about the target population; and the data analysis and statistics are the only keys to extract this knowledge from your underlying data.
Hence, devote enough time and energy to it and try to extract the most accurate and reliable results from your analysis. Otherwise the whole process of your surveying could be a waste.