User research: How to conduct in-depth analysis of multiple topics with a sum nature

Article Description: discussion on the analysis method of multiple topics in the survey questionnaire.

Introduction

In the quantitative study of questionnaires, in order to understand more comprehensively the research content, more extensive collection of information, often used to multiple-choice, but because of the multiple-directional features, in addition to frequency table and crosstab (only with single choice to do crossover), less use of other analytical methods, loss of a lot of useful information. In fact, if the research can make good use of multiple-choice, and in the analysis of the appropriate method, you can make full use of the information included in the multiple-choice, get more valuable conclusions.

in the last article, the paper discusses how to analyze the multiple topics with the classification in the questionnaire, this article will focus on how to analyze the multiple topics with the sum of the nature , and also can get the results of the group to follow the application.

Multiple selection with sum properties

in the questionnaire survey, often through multiple selection of users in a particular area of diversity, or awareness, ability, etc., at this point, the total sum of options can reflect the user in this aspect of the comprehensive situation, but also a way to analyze the problem.

In fact, with the sum of the nature of the topic, most of the problem is converted from the questionnaire, and the scale of the question of N-scale, which is equivalent to each option is a topic, in order to save the user's fill in the cost, commonly used multiple-choice alternatives. Of course, the sum of the relevant options of the questionnaire, more can reflect the user in a certain aspect of the comprehensive level, this does not repeat.

multiple topics with a sum of character generally have the following characteristics:

1, the title itself has the value of summation , such as the seller commonly used Third-party service tools (to investigate the diversity of tool applications), sellers daily use of security measures (to investigate the prevention of awareness), the user's commonly used network functions (to investigate the network use ability), The total score of the option summation of this kind of question has some reference value for the comprehensive analysis problem.

2, the options covered more comprehensive , the option should include as far as possible the main details of the investigation, if the area of investigation, the scope is relatively narrow, it is best to exhaust the main details; If the scope of the study is relatively wide, it is best to be able to make it to the whole, Divide into several levels and select the main detail points one by one.

3, the option is generally more than 5 , too few options are not necessary to sum, only more than a certain number, sum to have value, sum is to be summed up in the way, from the overall cognitive perspective, the general interpretation of user behavior or psychology.

4, usually not open, if the selection, SUM, total value gap is not easy to find samples in the study of differences, it is difficult to achieve the purpose of comprehensive evaluation.

5, the data is 0/1 format , that is, each option is a variable, selected for 1, not selected for 0, multiple choice to convert to this format, appropriate to do in-depth analysis.

When you have multiple topics with a summation, you can use the methods described below for in-depth analysis.

In this paper, two network questionnaires, respectively, is the last quarter of the "Seller Security Needs Survey", the study model with a multiple-choice, examined the seller's safety protection measures awareness; Four years ago, when I wrote the thesis, "The Relevance of life forms to instant messaging," The basic topic uses a multiple-choice topic, has inspected the netizen to the network use ability.

Sum directly

the direct summation method is very simple, because the multiple-choice data is 0/1 format, directly plus and after, can be expressed in the study, sample total score, reflecting the comprehensive level.

Sellers Security Protection awareness cases are slightly more complex, because the options contain mutually exclusive relationships, and the frequency of each option is as follows:

sum, you cannot add directly, you need to deal with it, because "there is no preventive or emergency measures" and other answers mutually exclusive, there are two ways to deal with:

1, will "no prevention or emergency measures" answer 0/1 do reverse processing (Recode processing: 0 to 1, 1 to 0), after all the options plus, if the total is divided into 0, it means "Safety prevention awareness" for 0, if the total score is greater than 0, It indicates that some measures have been taken to reflect the degree of safety prevention consciousness;

2, according to the situation to calculate the total score , if "no preventive or emergency measures" were selected, the option Plus and the variable is directly 0, if "no preventive or emergency measures" has not been selected, the remaining options plus and get the total score. At this point, the option Plus and the results are the same as Method 1.

In this example, the seller's overall safety prevention awareness average of 2, combined with the distribution of total score can be seen, the seller's safety prevention awareness is not strong, the means is relatively single, higher risk.

after cross analysis with other variables, we can also analyze the differences in safety prevention consciousness of each subdivision group.

Direct summation, in addition to the need to pay attention to the mutual exclusion between options, the most important thing is to determine whether the options for the investigation, is at the same level, approximately equal . In a case, the measure in each option represents a similar degree of safety prevention, not a particular measure that is difficult or difficult to achieve.

If each option is not at the same level and cannot be approximately equal in terms of inspection, other summation methods are required.

Weighted summation

to investigate the Internet users ' ability to use the network, if asked the interviewee directly, "What level of network use ability to achieve?" ", it may not be easy to obtain reliable data, you can indirectly use multiple-choice approach, ask respondents to choose to know how to operate the network applications, options should include the first, middle and advanced network applications, in order to fully evaluate."

because of the first, middle and advanced network applications, different levels, can not be approximately equal, can not be directly summed up, simple addition will cause great error, can not reflect the network use ability.

Therefore, the choice of reliability test, and then through the weighting method, the need for higher network applications to give a relatively large weight, the ability to require less than network applications to give smaller weights. Thus, the higher the total score obtained, the higher the network usage ability.

The weighted method used in practice is to convert the direct measurement x (score or proportion) into standard normal score Z and then to the weighted value W to obtain the measured value X of the comprehensive level.

In this example, the Cronbach ' alpha Reliability test of the network ability multiple-choice is obtained alpha=0.841, and then the sample ratio p is converted to normal normal score Z, and the weighted number is determined by the relative size of Z. Can calculate the network usage ability of each sample. The analysis shows that the average network use capacity of the sample is 27.9, and the standard deviation is 12.96. The calculation process follows the following table:

Use total score to group

After summing up the total score, you can group the samples, generally find high and low groupings, the comparison of high-low groups in other key variables, but also an analysis of the point of view.

There are usually two ways to group total scores:

1, the lower limit of the high group can be (mean value + 1 standard deviation); The upper limit of low grouping can be (mean value − 1 standard deviation), generally this method obtains the high and low grouping sample quantity, all does not necessarily reach the sample total 1/4, needs to judge whether according to the research goal is suitable. This approach is available when researching niche products.

2, another method, is the total score in ascending order, from the top down, in the sample total 27% of the location of the record sample score, as the upper limit of the lower packet threshold; from the bottom up, the sample score is recorded at 27% of the total sample, as the lower bound of the high group. The sample size obtained by this method can meet the needs of most research. This approach is available when researching mass products.

Summary

to sum up, with the sum of the multiple topics, the analysis of the process is as follows:

1, to determine the choice of multiple-choice summation, the total score whether there is interpretation of value, application value;

2, according to the characteristics of the choice, select the appropriate summation method: Direct summation, weighted sum;

3, the use of key variables (such as star, registration, etc.) on the overall grouping of samples, to see the breakdown of the number of different groups in the total score differences, in-depth analysis;
or using the total score to group the samples, focusing on the performance of the high and low groups on other variables.