# 1. Describe outliers and pick a technique that can be used t…

1. Describe outliers and pick a technique that can be used to handle outliers 2. Describe missing values and pick a technique that we have not discussed in this class to handle them 3. Discuss why and when do we need to use data normalization and standardization

**Answer**

1. Outliers are data points that are significantly different from the majority of the other data points in a dataset. They can arise due to various reasons, such as measurement errors, data entry mistakes, or genuine anomalies in the data. Outliers can have a significant impact on statistical analyses, as they can disproportionately influence the mean and standard deviation of a dataset, leading to biased results.

To handle outliers, one commonly used technique is the Z-score method. In this approach, each data point is transformed into a Z-score, which represents the number of standard deviations it is away from the mean of the dataset. Data points with Z-scores exceeding a certain threshold (often set as ±3) are considered outliers and can be treated accordingly. They can be removed from the dataset, replaced with more reasonable values, or analyzed separately, depending on the specific context.

2. Missing values refer to the absence or lack of data for one or more variables in a dataset. Missing values can occur for various reasons, such as non-response in surveys, equipment failures, or data corruption. Dealing with missing values is crucial as they can adversely affect data analysis and modeling.

One technique to handle missing values that we have not discussed in this class is multiple imputation. Multiple imputation is a statistical technique that involves creating multiple plausible estimates for missing values based on the observed data. This method takes into account the uncertainty associated with the missing values and provides more accurate estimates compared to single imputation techniques, such as mean imputation or regression imputation.

In multiple imputation, a model is built using the observed data to predict the missing values. This is done by using variables that are related to the missing variable as predictors. Multiple imputed datasets are then generated, where missing values are replaced with imputed values based on the model. The analysis is then performed on each imputed dataset separately, and the results are combined using statistical techniques that account for the variability between the imputed datasets, such as Rubin’s rules.

3. Data normalization and standardization are techniques used to transform variables to a common scale or distribution to facilitate comparisons and analyses. They are often employed in data preprocessing before applying various machine learning algorithms and statistical techniques.

Data normalization refers to rescaling the values of a variable to a specified range, typically between 0 and 1. This technique is useful when the actual values of the variables are not as significant as their relative positions within the range. Normalization can be achieved using techniques such as min-max scaling, where each value is subtracted by the minimum value of the variable and divided by the range (maximum value minus minimum value).

Data standardization, on the other hand, transforms variables to have zero mean and unit variance. This technique is beneficial when the variables have different scales and units, and we want to ensure that they contribute equally to the analysis. Standardization can be achieved by subtracting the mean of the variable and dividing by the standard deviation.

The choice between data normalization and standardization depends on the nature of the variables and the specific requirements of the analysis. Normalization is suitable for variables with bounded ranges, while standardization is often preferred for variables with unbounded ranges or when the scale differences between variables need to be accounted for.