The Complete Beginner's Guide to Discriminant Analysis in SPSS - No Math PhD Needed | Affordable Dissertation

Discriminant Analysis (DA), which is basically the statistical technique by which the observations or data get classified into different groups or categories, incorporating some particular variables called predictor variables. In simple cases, it predicts to which groups a new case would fall, the most likely given the relationship between the predictor variables and the most likely groups. 

For example, businesses can use DA to make predictions about whether a customer is likely to purchase a product based on their past purchasing behaviour, age, and income. In the field of education, DA can be used to create categories of students such as “high” or “low” performers, depending on how many hours they studied, how often they showed up to class, and how well they performed on their tests. 

Purpose and Use of DA in SPSS

This guide by Affordable Dissertation UK will introduce Discriminant Analysis in SPSS, which is a commonly used software tool and is used to make the process of this analysis a lot easier. The primary objective of this guide is to provide practical, step-by-step instructions to help users, such as beginners. 

You will learn how the statistical package for the social sciences (SPSS) can be used for the classification of your data, how to interpret the outcome and how you can use DA for your practical problem. The emphasis will be on getting some understanding of the concept and process of DA, without getting into detailed mathematical knowledge. 

SPSS Interface and Functionality

SPSS makes this process even easier with its easy-to-use interface, which comes equipped with some key sections such as Data View and Variable View. The Data View lays out the data, whether it is latitude and longitude or salesman name and sales amount. 

The data itself, with each row in the data representing a case and each column in the data representing a variable. The Variable View gives us great information about each variable (such as the variable type and label), which is important to ensuring we have the data in the correct format to perform our analysis on it. 

Learning Outcomes

Following this guide, you will gain a functional understanding of using Discriminant Analysis in the SPSS software without being required to have any advanced mathematical skills, and this will be a productive analytical tool for your data classification requirements.

Figure 1: SPSS interface highlighting relevant sections (Data View and Variable View)

Need expert support with your research? Our UK-based dissertation writing services can help you apply statistical tools like Discriminant Analysis in SPSS to your project with accuracy and confidence.

What Is Discriminant Analysis?

Discriminant Analysis is a statistical procedure of classified prediction, where, from a list of predictive variables, the selected variable is used to classify/predict a given observation in categories/groups. 

This is a particularly useful technique when the outcome you are seeking to predict is categorical, such as whether a customer will purchase or not (buy vs. not buy) or whether a person has a high risk for a particular health risk (high risk vs. low risk). By utilising historical data, DA creates a model that is able to categorise new data into previously defined classes with high accuracy.

Types of Discriminant Analysis

There are two main types of Discriminant Analysis:

• Linear Discriminant Analysis (LDA): LDA assumes that the predictor variables follow a normal distribution within each group and that the variance between the groups is equal. It is used when the groups are similar and the data points lie in a linear fashion. LDA is particularly useful when the data is well-behaved and the groups are relatively similar.
• Quadratic Discriminant Analysis (QDA): Unlike LDA, QDA does not assume that the variances are equal between the groups. This makes it possible for QDA to deal with situations where the groups are more distinct and where there is a different spread or variability of the predictor variables, and renders it rather flexible compared to LDA.

Similarities and Differences with Regression

Discriminant Analysis is often used to compare it to regression and logistic regression. Both methods are used for classification, but there are key differences between the two methods. 

While regression predicts a continuous result (as in 1 and 0) and Logistic Regression predicts a categorical result (as in yes/no or 0/1), Discriminant Analysis predicts the category of the dependent variable represented by continuous values. 

For DA, the predictors just have to be continuous and normally distributed; under logistic regression, categorical predictors can be multiplied.

Situations Where Discriminant Analysis Is Most Useful

Discriminant Analysis has a wide range of applications in domains such as marketing, healthcare, finance, and education. It is especially useful when you need to categorise data into several groups, based on continuous variables, such as predicting whether a person gets a loan (yes/no), based on their credit score and income, or classify whether people have a particular health treatment, based on medical test results.

SPSS Output for LDA and QDA

In practice, the decision between LDA and QDA is made on the basis of the structure of the dataset. To illustrate, consider the example of an output produced with the SPSS that compares the three models LDA, QDA, and Logistic Regression (LR) for the two-class classification of women as before and after statin treatment: The table shows that LDA had an accuracy of 86.9% on the training set and 79.8% under the leave-one-out condition, whereas in the case of QDA, the accuracy on the training set was similar (86.9%), but for cross validation the accuracy was less (65.5% in SAS). Logistic Regression outperforms both with an accuracy of 89.3%.

Figure 2: A comparison of LDA and QDA outputs with an example dataset

This example emphasises not only the functionality of DA, but also highlights the differences with other methods in practice. By inspecting outputs from this type of analysis in SPSS, like this, beginners can gain insight into the accuracy of classification by various techniques, highlighting the position of DA as a useful, interpretable, and practical method for the prediction of groups.

When to Use Discriminant Analysis in SPSS

Discriminant Analysis (DA) in the field of SPSS is a useful statistical procedure that conducts a classification of a categorical dependent variable by independent variables that are continuous in nature, and consists of two or more groups. It is ideal for situations where you wish to predict the group membership of an observation, such as responding whether a customer is going to buy a product (Yes/No) or grouping students into “pass” or “fail” categories based on their study habits, test scores or attendance.

In order for DA to work effectively in SPSS, some conditions need to be met. First, the dependent variable must be categorical, with two or more different categories of people, places, things, etc. For example, the result might be whether a person is in a high or low-income group. Second, the independent variables or predictive variables must be continuous rather than categorical, measurements such as age, income, test scores, etc., or measuring at an interval scale (e.g., ratio scale).

There are also some key assumptions to be aware of when using Discriminant Analysis in the Significance Passage of the data analysis in SPSS:

Normal Distribution of Predictors

The predictor variables should be distributed normally within each group. This assumption can be checked with histograms. A histogram where the distributions are bell-shaped curves for each group is normally distributed, and thus, this assumption is met. The histogram attached shows one way of proving this assumption by showing how a predictor variable follows a normal distribution, in this case, test scores.

Figure 3: A graphical representation of the Histogram showing the normal distribution assumption

Equal Covariance Across Groups

The variance of predictor variables should be roughly equal for all groups. This helps make sure that the model is not biased toward any group over another.

Independence of Observations

Each case should stand independently from any others. This means that the data from one observation should not have an effect on the data from another observation.

By making sure these assumptions are fulfilled and followed, Discriminant Analysis in SPSS can be a powerful tool for classification tasks.

Preparing Data in SPSS

The first thing you want to do is import the dataset into SPSS. It is very important to check whether there is a proper organisation of the dataset. The dataset must contain a categorical dependent variable (what we wish to predict) and continuous independent variables (what we use to predict the result). 

For example, if the variables are measuring a class of students and classifying them as “Pass” or “Fail” according to numbers of studying hours, the number of hours they spent studying and the test score, then the dependent variable will be categorical (for example, “Pass” vs. “Fail”), and the independent variables will be continuous (for example, achievable studying hours and test score).

Importing the Dataset

The first thing you need to do is arrange the dataset into SPSS. It is extremely important to check if the dataset is structured properly. The data set should include a categorical dependent variable (the outcome we want to predict) and continuous independent variables (the variables we want to use to predict the outcome). 

For example, suppose the problem is to classify students as “Pass” or “Fail” based on the number of hours of study and marks on a test, the dependent variable would be categorical (e.g., “Pass” vs. “Fail”), and the independent variables such as number of hours of study and test scores, would be continuous.

Coding the Dependent Variable

The dependent variable needs to be appropriately coded. In SPSS, categorical variables are usually coded as numerically coded series for the purpose of easy analysis. For example, “Group 1” might be represented as 1 and “Group 2” as 2 so that these categories can be easily understood by the program (SPSS) as two separate categories.

Handling Missing Values

It is important to determine whether the dataset has any missing values. SPSS has functions for handling missing data, either excluding cases that have missing data or imputing the missing values, whichever is appropriate based on how much of the data is missing. Proper handling of missing data is also important because this could otherwise negatively affect the results of the analysis.

Checking for Outliers

Outliers need to be identified and handled appropriately. Extreme values do have a strong impact on the accuracy of Discriminant Analysis, so extreme values must be addressed in order to get reliable results. 

A typical structure of a discriminant analysis dataset, with both the dependent variable and continuous predictors, is shown in the attached screenshot. By following the above-mentioned steps and ensuring proper data preparation, a solid foundation is laid for properly conducting accurate and reliable discriminant analysis in SPSS.

Figure 4:Example SPSS Data View showing the structure of a typical dataset for discriminant analysis

Step-by-Step Guide to Running Discriminant Analysis in SPSS

Running Discriminant Analysis (DA) in software such as, for example, the IBM program called ‘SPSS’ involves a series of well-defined steps to ensure that the analysis is done correctly. Here is a step-by-step guide to doing DA with screenshots and easy-to-understand instructions on each stage.

Step 1: Open Data in SPSS

To get started, open SPSS and open your dataset. SPSS will have the data available in two views: Data View and Variable View. The Data View is where the actual data is displayed in rows and columns. 

The Variable View is used to define the variables, for example, the type of data (categorical, continuous) and the label for the variable. The screenshot given explains how the data is represented in Data View, which shows the dependent and independent variables.

Figure 5: Open Data Example Dataset

Step 2: Navigate to Discriminant Analysis

Next, navigate to Analyse> Classify > Discriminant in the SPSS menu. This option opens the Discriminant Analysis dialogue box, where you can specify the variables for the analysis. The screenshot here shows the menu path, guiding users to the right location to access this function.

Figure 6: Menu Path

Step 3: Assign Variables

In this step, assign the categorical dependent variable (the outcome variable) to the Grouping Variable box. Then, move the predictor variables (independent variables) into the Independents box. If necessary, define the range for the grouping variable (e.g., Group 1 = 1, Group 2 = 2). The discriminant dialogue box screenshot shows how to assign these variables.

Figure 7: Discriminant Dialogue box

Step 4: Set Options

To configure the analysis, select the desired options under Statistics and Classify. In the Statistics section, choose Wilks’ Lambda and Box’s M Test to assess the assumptions of normality and equality of covariance. 

Figure 8: Checking the Required Summary Statistics

Under Classify, choose Leave-one-out cross-validation to evaluate model accuracy. The options window screenshot demonstrates these settings.

Figure 9: Classifying the Data Using Leave one out Classifies

Figure 10: Save Option that classifies the predicted Group

Step 5: Run the Analysis

Finally, click OK to run the analysis. SPSS will generate the output, which includes the results of the DA. The output window screenshot shows where to find and interpret the results, including classification accuracy and significance tests.

By following these steps, users can successfully perform Discriminant Analysis in SPSS, allowing them to classify data into meaningful groups based on predictor variables.

Figure 11: Save Option that classifies the predicted Group

Interpreting the Output

Once you have completed Discriminant Analysis in SPSS, it outputs show a variety of relevant tables that aid in the interpretation of the data results. Here is a breakdown of the most important components:

Eigenvalues Table

This table indicates how much variance each discriminant function explains. Higher eigenvalues suggest that the function does a better job of differentiating between the groups. The larger the eigenvalue, the better the discriminant function separates the groups based on the predictor variables.

Figure 12: Example illustration of the Eigenvalues Table

Wilks’ Lambda

This statistic tests the significance of the discriminant function. A smaller Wilks’ Lambda value indicates a more statistically significant discriminant function. If the value is close to zero, it suggests that the function is effective in distinguishing the groups.

Figure 13: Example illustration of Wilks’ Lambda

Standardised Canonical Discriminant Function Coefficients

 These coefficients show the importance of each predictor in classifying the cases. Higher absolute values of these coefficients indicate that the corresponding predictor has a greater impact on the classification process.

Figure 14: Standardised Canonical Discriminant Function Coefficients

Classification Results Table

This table describes how well the model predicts the classifications of new observations. It offers accuracy measures, such as how many times the model correctly classifies a case (i.e., what percentage of cases is correctly classified). 

Figure 15: Classification Results Table

Provided below is the screenshot, which shows the above-mentioned key tables in order to visualise the SPSS output and get an understanding of the output results.

Example Scenario

In this example scenario, let’s say we are using Discriminant Analysis to predict student performance (categorised as High vs. Low) based on predictors like study hours, attendance, and test scores. After running the analysis in SPSS, we obtain the output, which includes important tables like the Classification Results Table.

Using the example table that I attached below, the Classification Results Table gives us the original and predicted group membership for each case. For instance, if a student was originally categorised as “customer service,” the table shows how well the model predicted this classification. It also shows the percentage of students from each group that were correctly classified into their predicted groups.

Figure 16: Example case Scenario

From the output:

• Customer service students were correctly classified 82.4% of the time.
• Mechanic students had a lower correct classification rate of 17.2%, and dispatch students were predicted correctly 77.3% of the time.

This information helps us evaluate the performance of the discriminant function. The Classification Processing Summary indicates how many cases were included in the analysis (244 cases), and the Prior Probabilities for Groups table gives the distribution of students across the categories (each group has a prior probability of 0.333).

Common Mistakes to Avoid While Doing Discriminant Analysis in SPSS

When using the Discriminant Analysis procedure in IBM SPSS, it is necessary to be aware of typical errors when applying the analysis that may cause inaccurate results. Here are a few to avoid:

Using Non-Continuous Predictors

 Discriminant Analysis works well when comparing continuous predictors. Using categorical variables, such as whether a person is male or female or has a high school education or a college degree, can negatively affect the results. To accurately classify, make sure that your predictor variables are numeric and on an interval or ratio scale.

Ignoring Multicollinearity Among Predictors

Multicollinearity is a phenomenon that happens when predictor variables are highly correlated with one another. This may lead to inflated standard errors and make it challenging to identify the impact that each predictor has on the outcomes. One of the first steps to running this analysis is to check for multicollinearity using tools such as correlation matrices or the Variance Inflation Factor (VIF).

Forgetting to Check Assumptions

Before conducting Discriminant Analysis, always verify the important assumptions, especially normality of the predictors, equal covariance between the groups, and independent observations. Ignorance of these assumptions can result in faulty conclusions. For example, using data that are not normally distributed without performing some sort of coding can impact the accuracy of your results.

Overinterpreting Results When Classification Accuracy is Low

 While it is desirable to pursue high classification accuracy, low accuracy is not necessarily an indicator that the model is ineffective. It might mean that the data is complex or the model is too simple to learn. Always look at the results holistically, taking into consideration both statistical significance and model fit.

Best Practices for Conducting Discriminant Analysis in SPSS

When conducting DA in IBM’s SPSS, there are a few best practices that should give you an advantage in ensuring that your results are reliable and meaningful. Here are some of the most important practices that you must consider:

Validate the Model

Validation should always be done on the model (either cross-validation or holdout sample validation). Cross validation-a method where subsets of the data are used for training and testing to help evaluate how well the model is expected to generalise to new data. 

This is a crucial step to ensure that the model is not overfitting the data, which means that the model is not just memorising the training data but can make accurate predictions on new, unseen cases.

Report Key Metrics

When reporting the results, it is important to report important statistics, such as classification accuracy, Winks Lambda, and how important the predictor variables were. Classification accuracy refers to the degree to which the model has been able to place observations in the correct groups. Wilks’ Lambda test for the significance of the discriminant function, and knowing which predictors make the biggest difference, helps in interpreting the analysis.

Use Visualisations for Clarity

Visualising the results can enhance understanding and presentation. Plots such as group means plots or classification boundary plots can help illustrate how well the model discriminates between groups. These visualisations make it easier to interpret complex results and communicate findings effectively. The attached graph demonstrates the use of such plots for better clarity.

Figure 17: Example of group means plots or classification boundary plots

By following these best practices, users can ensure that the results of their DA are accurate and interpretable.

Conclusion

Discriminant Analysis is a significant and convenient classifying tool, which assists researchers in categorising data related to continuous predictors. While the actual statistical theory on which it is based may seem very complicated, the reality is simpler and easier when using SPSS, so that even the most inexperienced people (i.e. without advanced mathematical skills) can easily get started. 

By following along with the step-by-step guide given here, users can easily perform Discriminant Analysis and obtain meaningful insights from their data. SPSS simplifies the whole process so the researcher can concentrate on interpretation instead of complicated calculations. 

For people unfamiliar with Discriminant Analysis, it is good to practice with some sample datasets to gain confidence and know how to effectively apply the technique. With continued practice, beginners can confidently apply the Discriminant Analysis technique to classify data in real-world research applications, from market segmentation to medical studies, and improve their analytical skills and research outcomes.

If you are struggling to apply these methods in your own research, connect with our highly qualified dissertation writers online. They can guide you step by step and ensure your data analysis is accurate and well-presented.

FAQs 

What is discriminant analysis used for in research?

Discriminant Analysis (DA) is used to classify cases into predefined groups based on predictor variables. It is widely applied in research for predicting categorical outcomes, such as identifying whether a customer will buy a product, categorising patients based on risk factors, or predicting student performance based on study habits.

How does linear discriminant analysis differ from logistic regression?

While both Linear Discriminant Analysis (LDA) and logistic regression are used for classification, LDA assumes that the predictors follow a normal distribution and have equal variance across groups, whereas logistic regression does not make these assumptions. LDA is often preferred when dealing with normally distributed data, while logistic regression can be used when data does not meet these assumptions.

When should you use discriminant analysis instead of regression?

Discriminant Analysis is typically used when the dependent variable is categorical with two or more groups, whereas regression is used when the dependent variable is continuous. DA is preferred for classification tasks, while regression is used to predict a continuous outcome.

How do you interpret coefficients in discriminant analysis?

The coefficients in DA show the importance of each predictor in the classification process. Higher absolute values indicate that a predictor has a greater influence on the discriminant function. These coefficients are used to determine the boundaries between different groups.

What are the assumptions of discriminant analysis in statistics?

The key assumptions of DA are: normal distribution of predictors, equal covariance across groups, and independence of observations. These assumptions ensure that the model accurately classifies the cases.

How do you run discriminant analysis in SPSS step by step?

To run DA in SPSS, you first need to import the data, assign the categorical dependent variable and continuous predictors, set the analysis options (such as Wilks’ Lambda and cross-validation), and then run the analysis. SPSS will generate output with the results.

What are real-life applications of discriminant analysis in business?

In business, DA is used for customer segmentation, market research, fraud detection, and predicting customer behaviour. It helps businesses classify customers into different groups, such as high vs. low value, and make informed marketing decisions.

Can discriminant analysis be used in machine learning?

Yes, Discriminant Analysis can be used in machine learning, especially in supervised learning tasks where the goal is to classify data into categories. LDA and QDA are often applied in pattern recognition and classification problems in machine learning.

How accurate is discriminant analysis compared to other classification methods?

Discriminant Analysis can be very accurate when the assumptions of normality and equal covariance are met. However, its performance may decline when these assumptions are violated. Compared to methods like decision trees or support vector machines, DA may be less flexible but is still widely used for classification tasks when the data fits its assumptions.