CH4-SPSS Statistics Operation Advanced

Knowledge Points:

Single-choice, multiple-choice input

Data cleansing prior to analysis, including deletion of duplicate records, outliers, logical checks

Data weighting in the case of inconsistent population and sample distribution structures

Cross-table when analyzing related issues

1. Project background

2. Questionnaire input

The definition of a single choice:

When the value of the variable is defined, such as: 1 = "male", 2 = "female", in the input can be entered by the "Show Finger Label" icon to select the method "select male or female" to enter.

Definition of multi-choice:

First determine the use of binary or multiple classification of the input;

Second, there are 2 ways to define a set of multi-option variables, one in the Multiple responses submenu and one in the Table submenu. The former belongs to the base module, the corresponding settings can not be saved, can not be used in the Tabulation module, the latter belongs to the table module, and can be saved in the data file reuse, can be used in the process of ownership table.

Here's how to use the second method:

3. Questionnaire quality Check

To remove duplicate records:

Step one: Find duplicate records, step two: Process the duplicate records.

Step One:

Step Two:

Outliers found:

For each variable has a certain range of values, due to input errors caused by the outliers to be processed before the analysis.

By observing the frequency distribution of the variable (analysis-description statistics-frequency) to identify outliers, and "Select cases" to remove outliers.

Logical check:

Such as: personal income is greater than household income, unmarried but have children, etc. are contrary to logical common sense.

Step one: see if there is a logical problem (analysis-table-Set table) by tab

Step two: Dealing with cases that are illogical. Delete the words in the "data-Select Case" process.

Step One:

4. Questionnaire Data analysis

Weighted by questionnaire:

When to be weighted: the distribution structure of the sample is inconsistent with the overall structure. Such as: the overall ratio of male to female is 6:4, but the male and female ratio of the sample questionnaire is: 7:3, this time needs to give a certain weight to the questionnaire data.

Weighted thinking: Determine the variables that affect the results-calculate the number of samples under the variable ratio weight-based on weight adjustment case

Step one: On-demand weighted variables into subtotals, and save as new files.

Step two: Under the new file, calculate the total number of samples, and then calculate the various types of the proportion. That is: the number of categories/total samples.

Data-subtotals

Convert-Calculate variable

Step three: Manual input of the overall proportion, namely: the overall gender distribution and education level distribution. Then calculate the weight = overall ratio/category proportion.

The calculation weights are in the "convert-Calculate variable".

Step four: Combine the weights into the questionnaire data. Prior to this, the questionnaire data needs to be sorted by these two variables.

Sort:

File Merge:

Step five: Add the weight variable after the questionnaire data, to be based on this weight data weighting.

The weighted data will have the word "weighted range" in the lower-right corner of the data view.

Business Analysis:

Do various cross-chart business analysis, such as analyzing the distribution of payment methods in gender, and the impact of gender on payment methods.

The frequency distribution of univariate variables is done, and the related variables are cross-linked.

Analysis-table-Set table

CH4-SPSS Statistics Operation Advanced