Machine Learning Using R
This course is an advanced course designed by the integration of Research Methodology and Applied Statistics Using R. Complete the courses on your own to understand this machine learning course to a greater extent.
Introduction Slides
Hypothesis test
Hypothesis is a scientific guess which has not tested yet. Hypothesis test is done to make decision on the nature of a variable or statistics. Before performing a test we need to make a null hypothesis (H0).
Example:
H0: Two means are equal (t-test)
H0: All means are equal (ANOVA)
H0: Two variances are equal (Levene test)
H0: Correlation coefficient = 0
H0: Regression coefficient = 0
H0: No association between the variables (chi-square test)
H0: No effects/relationships (correlation, regression, ANOVA)
H0: Data is normally distributed (Shapiro test)
Look at this applet https://kamrulext.shinyapps.io/p-values/
p-values indicate evidence to support the H0.
p-value shows level of error (type I: true negative) in repeated trials.
FYI, type II error = false positive (fails to reject a false H0)
p-values remain in the rejection area.
p-value is the area under the curve left side of the z-value.
Each of the p-values correspond to a z-value.
Z-value => quartile value, variable value, observed value, x-axis value, test statistics (t, r, F, beta).
H0 is born to be rejected, so, we can never accept it. We can only reject or not reject it.
So, in a simple sense, small p-value (<= 0.05) means insufficient evidence to support the H0 (rejected).
Large p-value (>0.05) means sufficient evidence to support the H0 (cannot be rejected).
Here is the twist: H0 rejected => significant relationship or effect because H0 was saying no relationship
Further twist: Test statistic (t, r, F, beta) value increases p-value decreases.
Higher test statistic gives small p-values => significant result.
Test statistic equivalent to (Coefficient)/(Standard error)
Schedule and data sets
| Lesson | Topic | Duration | Links |
|---|---|---|---|
| 1 | Installation of R and RStudio |
1 hr | Instruction, YouTube |
| 2 | Introduction to R |
3 hr 30 min | YouTube, R Codes Used in the Class |
| 3 | Tidy Thinking in R |
3 hr 30 min | R Codes Used in the Class |
| 4 | Linear Regression |
3 hr | R Codes Used in the Class |
| 5 | Bias-Variance Trade-Off |
2 hr | R Codes Used in the Class |
| 6 | Step-wise regression |
1 hr | R Codes Used in the Class |
| 7 | Ridge regression |
1 hr | R Codes Used in the Class |
| 8 | Lasso Regression |
0.5 hr | R Codes Used in the Class |
| 9 | Validation: Train-Test Split |
1 hr | R Codes Used in the Class |
| 10 | Cross Validation |
1 hr | R Codes Used in the Class |
| 11 | Bootstrapping |
0.5 hr | R Codes Used in the Class |
| 12 | Logistic Regression |
2 hr | R Codes Used in the Class |
| 13 | Moving Beyond Linearity (Polynomials) |
3 hr | R Codes Used in the Class |
| 14 | K-Nearest Neighbors (KNN) |
3 hr | R Codes Used in the Class |
| 15 | Linear and Quadratic Discriminant Analysis |
1 hr | R Codes Used in the Class |
| 16 | Support Vector Machines |
1 hr | R Codes Used in the Class |
| 17 | Random Forest |
1 hr | R Codes Used in the Class |
| 18 | Gradient Boosting |
1 hr | R Codes Used in the Class |
| 19 | Neural Network |
2 hr | R Codes Used in the Class |
| 20 | Principal Component Analysis |
2 hr | R Codes Used in the Class |
| 21 | Cluster Analysis |
2 hr | R Codes Used in the Class |
| 22 | Time Series Analysis |
2 hr | R Codes Used in the Class |