What is bivariate analysis – Bivariate analysis is one type of analysis used by the number of variables. This simple analysis is capable of producing very useful tests and statistical model
In general, there are 3 types of variable:
1. Univariate analysis (1 variable)
2. Bivariate analysis (2 variables)
3. Multivariate analysis (more than 2 variables)
Every kind of variable has its own unique criteria and analytical tools. Knowing a complete bivariate analysis will make it easier for you to determine the type of analysis that is accurate.
What is Bivariate Analysis?
Bivariate analysis is an analysis that is performed to determine the relationship between 2 variables. In this analysis, two measurements were made for each observation.
In this case, the samples used could be pairs or each independent with different treatments.
In general, in a bivariate analysis, the variables used can be related or independent (independent). Interrelated means that the same sample is given 2 different measurements.
Meanwhile, independent means that measurements are made on two different sample groups.
Types of Bivariate Analysis
I. Descriptive Analysis
In the descriptive analysis, bivariate analysis can apply to almost all data visualizations. Types of visualization displays such as bar charts, line charts, column charts, etc. can still be used for bivariate analysis.
One of the interesting data visualizations that are usually done with bivariate analysis is a scatterplot.
The scatterplot is a data visualization in the form of points displayed on the x and yes axes. The x and y axes represent the value of each variable.
By using a scatterplot, we can see the pattern of the relationship between the 2 variables. The relationships that are formed can be linear, exponential, seasonal, etc. according to data conditions.
Don’t forget, the scatterplot is only a tool to detect relationship patterns, not to draw conclusions on the relationship pattern between 2 variables.
II. Inferential Analysis
By using inferential analysis, you can draw valid conclusions in testing 2 variables.
Speaking of inferential analysis, there are many types of statistical tests that you can do with 2 variables.
Here is a little list of types of test analysis that you can do:
1. McNemar test
The McNEmar test is a bivariate test used to test before and after treatment (Pre-Test and Post-Test) where each individual is used as his own controller.
This test is performed for nominal and ordinal data measurement.
This test is used to test the effectiveness of a particular treatment on sample conditions.
For example, this test is used to determine the effect of moving a person from rural to urban areas to political preference.
2. Sign Test
The Sign test is used to determine whether there is a difference between ordinal data obtained from the same sample and pairs.
The thing to remember from the Sign test is that this test is only able to determine whether there is a difference, not the size of the difference.
This test is done by giving a positive or negative sign of the difference between data pairs.
Sign tests can be used to identify a person’s tendency towards 2 product brands.
The data scale used in this test is ordinal
3. Wilcoxon Matched Pairs Test
The Wilcoxon test is a test conducted to determine whether there is a relationship between two variables or not.
The data scale used in this test is ordinal.
4. Paired t-test
The paired t-test is a two-variable test conducted to determine whether there is a significant difference in the mean or not.
An example of using a paired t-test is testing whether there is a significant difference in the average between mathematics and art scores of grade-A students.
5. Fisher Exact Probability Test
The Fisher Exact Probability test is a test conducted to determine the significance of a comparative hypothesis in two small samples. independent.
This test is used when the data conditions are nominal and ordinal.
In the calculations, the data in this test are grouped into 2 independent groups. For example, men and women, then the poor and not poor.
Later, these calculations will be grouped into a 2×2 contingency table.
6. Chi-Square two sample test
Chi-Square two-sample test is used to determine whether there is a relationship between the 2 variables or not.
In the two samples chi-square test, the data scale used was the nominal scale.
7. Median Test
This test is used to test the comparative hypothesis of two independent samples. In this test, the data scales used are nominal and ordinal.
This test is based on the median sample taken at random.
The data scales used in this test are nominal and ordinal.
8. Mann-Whitney U-Test
Mann-Whitney U-Test was used to determine the significance of the differences between the two populations.
In this test, the data scale used is ordinal.
An example of the Mann-Whitney U-Test test is a teacher who wants to find out whether students in their class have talent in mathematics or are more dominated by tutoring assistance.
9. Kolmogorov Smirnov Test
The Kolmogorov Smirnov test is a test conducted to determine whether two variables have the same distribution or not.
This test is commonly used to prove whether the two variables used come from the same distribution before further analysis is carried out.
The data scale used in this test is the interval and ratio.
10. Wald-Waldovitz Test
The Wald-Waldovitz test is a test that is carried out whether the two variables used to come from the same population or not.
In this test, at least the data used has an ordinal scale.
11. Independent t-test
The Independent t-test is a test conducted whether 2 variables from different groups have the same mean or not.
In this test, the data scales used are intervals and ratios.
For example, a researcher wants to prove whether the average score of the final exam for a favorite school is significantly different from that of a non-favorite school.
12. Correlation Analysis
Correlation analysis is an analysis used to determine the relationship between two variables. With correlation analysis, we can find out whether 2 variables have a positive or negative relationship.
It is important to remember that correlation is simply an analysis that explains how strong the relationship between 2 variables is.
Correlation analysis cannot be used as a basis for concluding a causal relationship between 2 variables.
An example of using correlation analysis is the relationship between student height and weight.
13. Simple Linear Regression Analysis
Simple linear regression analysis is an analysis used to determine the effect of a variable on other variables.
In contrast to correlation analysis, simple linear regression analysis aims to explain the causal relationship (causality) between the independent variables and the dependent variable.
With this analysis, we can conclude to what extent one variable affects other variables.
Example of Using Bivariate Analysis
A researcher wants to know how the relationship between the weight and height of school students. Based on a sample of 60 students, the following data were obtained:
Based on the data above, perform descriptive and inferential analysis.
Based on the data above, there are several things we can do as initial identification to carry out further analysis:
1. We use 50 data. Based on the central limit theorem, we can assume that the data is normally distributed.
2. The data used has a ratio scale.
3. There is 1 sample group with 2 different types of data (paired).
Based on the identification results above, we can perform descriptive analysis in the form of frequency distribution, data visualization in the form of a scatter plot, and inferential analysis using paired t-test.
For the use of a scatterplot, you can follow these steps:
1. Graphs> Legacy Dialog> Scatter Dot
2. Click simple scatter> define
3. Enter the variables to be made in the scatterplot on the x and y axes
4. In the menu titles, type the scatterplot title that we want. want, click continue
5. Click Ok
The following is the scatter plot diagram results based on the data we use.
Based on the picture above, it doesn’t really look like the pattern of the relationship between height and weight. indirectly, we do not see a strong correlation between the two variables.
To confirm this, we can use correlation analysis and see how strong the relationship between student height and weight is.
For correlation analysis, I won’t go into the tutorial too much. You can read a very complete correlation analysis article that I have written on this blog.
Based on the results of the correlation analysis, it appears that the value of r = 0.299. It can be concluded that there is a weak positive linear relationship between student height and weight. The taller a student is, the greater the weight value he has but only in small terms.
To ascertain whether height affects student weight, we can use simple linear regression analysis.
1. Analyze > Regression > Linear
2. Input the variables that we will analyze. In this case, weight is the dependent variable while height is the independent variable.
3. Click Ok
4. Here is the output we get
Based on the test results, there are 3 main things that we conclude:
1. Height affects the weight of students by 0.90 percent. This can be seen from the r square value of 0.90.
2. The model formed based on the results of the above analysis is y = -61.197 + 0.788x
Based on this equation, it can be concluded that every 1 cm increase in the student’s height, the student’s weight will increase by 0.788 kg.
3. Based on the results of the significance test, it can be seen that the p-value is bigger than alpha (0.05). This proves that there is no significant positive relationship between height and weight. We have not enough evidence to prove that the higher a student, the weightier they are.
The bivariate analysis is an analysis conducted on 2 variables.
In bivariate analysis, a researcher can apply descriptive analysis and inferential analysis.
The descriptive analysis which is interesting to use in 2 variables is the scatterplot. For inferential analysis, various types of tests can be performed depending on the type and scale of the data.
By knowing the types of bivariate analysis, you can come up with a systematic and valid conclusion.