A Quick Guide for Advanced Features
In this guide, we only explain some key concepts and demonstrate advanced analytical features. The following example survey shows how AllCounted empowers users like you in an innovative and unparalleled way.
Suppose that you are working on a survey project with two goals:
- Understand the relationship between students' sleep and their academic performance
- See if gender or sports activities play any role
Let's assume that you have the following four questions in your survey:
Question 1 How many hours do you sleep on average, on a nightly basis? Please enter a number only.
Question 2 What is your cumulative GPA, on a 4.0 scale? Please enter a number only.
Question 3 How much time do you spend on sports on a weekly basis?
- Less than an hour
- 1 - 2 hours
- 2 - 4 hours
- 4 - 6 hours
- More than 6 hours
Question 4 What is your gender?
In this guide, you'll accomplish the following tasks:
- Create a survey with questions that collect quantitative, categorical, and ordinal data
- Conduct two-way cross tabulation between student gender and time spent playing sports
- Do a student's t test to see if GPA means are statistically different between female and male students
- See if gender and time spent playing sports are associated in any way
- Perform linear and polynomial regression between time spent sleeping and GPA
Questions and Data Types
Each question collects some data from survey participants. The type of data a question collects determines the analysis available on the question. AllCounted supports three data types:
|Categorical||Example– Hair color: Black, Brown, Blonde, etc.||Percentage, cross tabulation, Chi-square test of independence|
|Ordinal||Example – Strongly agree, Agree, Neutral, Disagree, Strongly disagree||Percentage, mean, median, mode, cross tabulation, Chi-square test of independence|
|Quantitative||Examples – Age, income, blood pressure, temperature, number of people in a family, balance on a bank account, etc.||Mean, median, mode, standard deviation, t tests, ANOVA, linear or polynomial regression|
Ordinal data is a special kind of categorical data. We make this distinction only when needed. Otherwise, they are both referred to as categorical data.
For Questions 1 and 2, you want to collect numbers. To enforce this requirement and make analysis on quantitative data possible, you have to indicate that the data type of these two questions is quantitative. Also, we highly recommend that you use validation to require survey participants to only enter numbers (with or without range requirement) because invalid data will be excluded in the analysis. See the following two screenshots:
Suppose that you select Ordinal as the data type for Question 3 and Categorical for Question 4. Please see the following two screenshots:
Campaigns, Distributions, Analysis Groups, and Filters
We've made great efforts to give survey creators maximum flexibility in handling their surveys and results. At AllCounted, a survey is considered as intellectual work, which is basically a list of questions. The first step in making a survey available to participants is creating a campaign. A campaign decides when a survey is live and who can take it. Once a survey becomes live, you need to choose how to distribute your survey to people, and the options include web links, emails, Facebook links, and embedding in a website. This three-level survey-campaign-distribution structure creates maximum flexibility in targeting survey participants and analyzing responses.
A survey can have multiple campaigns that target the same or different groups of people at the same or different times. Each campaign can have multiple distributions to one or multiple groups of participants. We let you organize survey data at the survey, campaign or distribution level and to analyze survey responses separately, as a whole, or in any possible combination. Additionally, you can compare results between campaigns or distributions through cross tabulation.
For greater flexibility, we allow you to put certain survey responses into one or more labeled groups, which we call Analysis Group. With this tool, you can analyze results by group, or you can even compare results between groups through cross tabulation.
Another benefit of campaigns, distributions, and analysis groups is that they can be used as data groups in statistical analysis. When a statistical tool needs more than one data group, not only can you use the answer choices of a Multiple Choice question to split responses and create groups, but also campaigns, distributions, or analysis groups. Moreover, campaigns, distributions, and analysis groups can be treated as a categorical variable used in a Chi-Square test of independence and other statistical tests.
In addition to campaigns, distributions, and analysis groups, you can always use individual questions and information about responses as conditions to slice and dice responses into whatever datasets you'd like for analysis (Tip: You can use Analysis Group or Shortcut to remember a desired dataset). All these conditions are collectively called Filters. The following screenshot shows these filters:
Suppose that you have collected enough responses and are ready to perform analysis. Open your survey, click Analyze, Results as shown below:
The above screenshot shows the types of responses you can use in analysis and the available analytical tools (under View). You can look at and analyze the responses to your survey in different views. The default and perhaps the most important view is Summary. In the Summary view, AllCounted lists the text answers to a Text Box question by default. If the question's data is quantitative, you'll see a Statistics link, which gives you some simple statistics on the responses to your question (Note that you can get many more statistics under the Statistics view). Please see the following two screenshot shots:
In the Summary view, for Multiple Choice questions like Question 4, AllCounted displays the statistics in a chart and table as shown below. The chart and the table have the same information, but the table is helpful if the text is too long to display on the chart.
Question 3 is special because its data is ordinal. In the Summary view, you can see more statistics such as the mean, median, and mode. Please see the following screenshot. Note that we calculate the mean for ordinal data for practical reasons. We automatically assign 1, 2, etc. to the first row (Less than an hour), second row (1 - 2 hours), etc. when calculating these statistics.
Understand Your Data First
Before you start doing advanced analysis, it's always good to have a basic understanding of the responses to the question you want to analyze. Another reason to review the responses first is because some statistical tests assume that data has met certain requirements; otherwise, an analysis could be meaningless. For example, student's t tests assume that data is normally distributed. A basic understanding of your data can help prevent this type of error.
For quantitative Text Box questions, we offer a wide range of basic statistics such as quartiles, central tendency, dispersion, and distribution. You can even split the data into a number of groups of equal percentile or find a specific percentile. In addition, you can create histograms or boxplots for quantitative responses. To do this basic analysis, click Analyze, Results, and then Exploring Quantitative Data under the Statistics view. You'll be asked to select the variable to analyze. The following two screenshots show this type of analysis on Question 2 (the GPA question).
You have four ways to do cross tabulation: by the answer choices of a Multiple Choice question, campaigns, distributions, or analysis groups. Please see the following screenshot:
For example, let's select "By a Multiple Choice Question" and click the Apply button. You'll be prompted to select a Multiple Choice question. Click the Question menu and select a Multiple Choice question. Suppose we selected Question 4 and its two answer choices, as shown below:
Click the Apply button, and you'll see the data separated by the answer choices of the selected Multiple Choice question.
The data in the above screenshot is equivalent to the following two tables (The second table is only applicable to Multiple Choice questions with the ordinal data type.):
|Less than an hour||1 - 2 hours||2 - 4 hours||4 - 6 hours||More than 6 hours|
|Female||1, 11.11%, 5.88%, 5.88%||2, 22.22%, 11.76%, 11.76%||3, 33.33%, 17.65%, 17.65%||2, 22.22%, 11.76%, 11.76%||1, 11.11%, 5.88%, 5.88%|
|Male||1, 12.50%, 5.88%, 5.88%||3, 37.50%, 17.65%, 17.65%||4, 50.00%, 23.53%, 23.53%|
|Mean, Median, Mode|
|Female||3.00, 3, 2 - 4 hours|
|Male||3.25, 3.5, 4 - 6 hours|
If the data type of a Text Box question is quantitative, then you'll see how its data differs between answer choices. The following screenshot shows how the data of Question 1 varies depending on the answer choices of Question 4.
The data in the above screenshot is equivalent to the following table:
|Mean of Sleep Hours|
Independent Samples Student's T Test
AllCounted supports one sample, independent samples, and paired samples student's t tests. In this guide, we use the independent samples t test as an example. Suppose that you want to know whether the difference of GPA means between female and male students is statistically significant. Click Independent Samples Student T Test under the Statistics view and click the Apply button. You'll be prompted to select the variables and data groups. Here are our selections for this demo:
The following screenshot shows the results, which indicate that the difference in GPA means is not statistically significant.
Here is a neat feature for survey creators. If you'd like to export the data points in this test and put them in another analytical tool, you can click on the name of a data group to show all the data points and copy all of them with a mouse click. This feature is available for all statistical tools. The following screenshot is an example:
It is worth noting that for the two data groups in this test, you can also select two campaigns, two distributions, or two analysis groups. This applies to all the statistical tools that need two or more data groups. Besides, you can always use the Filters to further segment responses into the dataset you'd like. The possibilities are endless!
Chi-Square Test of Independence.
To conduct a Chi-square test of independence, you need to have two categorical variables. In our example, the gender and time spent playing sports questions meet this requirement. Click Chi-Square Test of Independence under the Statistics view, then click the Apply button. You'll be asked to select the variables. Here are our selections for this demo:
The following screenshot shows the results based on the above settings.
Note that in this test, you are not asked to select any data group, which means that you have only one group, which is defined by your "Filter by" settings. for this and other tests such as the one-way ANOVA we offer, you can always use filters to slice and dice data, and the results will be updated instantly.
Doing one-way ANOVA is similar to doing an independent samples t test. The only difference is that you can select only two data groups for the t test, but you can select more than two groups in the one-way ANOVA test. To do one-way ANOVA, click One-Way ANOVA under the Statistics view, then click the Apply button. You'll be prompted to select two or more data groups and the dependent variable. The following two screenshots show our selections in this demo and the results:
Note that our system is very flexible and you can select any two or more answer choices to form the data groups needed for this test.
Regression can be performed between two Text Box questions with the quantitative data type. To perform regression between Questions 1 and 2, select "Simple Linear Regression" or "Polynomial Regression" under the Statistics view and click the Apply button. You'll be prompted to select the independent and dependent variables. Let's select Question 1 as the independent variable. Click the Apply button, and you'll get the following table, function, and chart:
For polynomial regression, you can optionally specify the degree of the fitting polynomial function. You may enter a number between 1 and 20 for the degree. The higher the degree, the more time it'll take to produce the regression function and chart. If you don't select the degree, we'll use 2 by default. The following screenshot shows this:
If our analytical tools can't satisfy your needs, you can export all or some of your survey responses into a text, xml, or Excel file. Click the Results menu, select Individual Responses, and click the Apply button, which will display a list of individual responses. Then, click the Export menu and the download format. The download will start immediately. Please note that if you have a lot of responses, the download may take a while. The good news is that you can leave the download window open and start another browser window to continue using our website without waiting for the download to finish.
As mentioned above, in statistical tools, you can click the name of a data group and export all the data points used in a test.
Let's Help Each Other!
We offer many powerful and innovative tools for online surveys, and you need to sign up to use them. All tools and services at AllCounted are completely free. If you have any questions, suggestions or see something that's unclear, missing, or incorrect, please don't hesitate to contact us at firstname.lastname@example.org. We will do our best to timely serve your needs.
If your colleagues, friends, or classmates need a tool for online surveys, please let them know that we can help. Thank you!