Question

Discuss in detail the pros and cons of using transformations. Also discuss any alternatives to transforming your data.

Answer 1

The transformation helps in organized the data in a better way but it leads to computational burden which slows down the process.

Non-linear model and weighted least square linear regression are the alternatives of transforming data.

Pros of data transformation

i. It improves the quality of the data and protects from the variations resulting from data mines such as null values, wrong indexing, duplicated data or incompatible data formats

ii. It ensures compatibility between systems, applications and data types which enables multiple use of data.

iii. It enables better organization of data for easier use by humans and systems.

iv. It improves data normality and homogeneity

Cons of data transformations

1. Problems may arise during transformation due to carelessness or lack of expertise

2 It may be very costly to transform data due to specialized infrastructure needed such as software, tools or skilled person

3. Needs may change after data transformation that do not suit the transformed data and may need reverting to original data.

4. Transformed data may achieve normality test but may not reflect the true trend of original data analysis

5 Organized the data in a better way

6 Improves the data quality

7 Correct the problem of non-normality and unequal variance.

Explanation:

Alternatives to transforming your data.

Alternatives to transforming data are:

1.    Try different linear model

2.    Use nonlinear model

3.    Weighted least square linear regression

4.    Alternative straight line regression methods

5.    Removing outliers.

a) Non-parametric tests

This is a test that uses non-normality tests without making usual distributional assumption of normally distributed data. An example is using Kruskal-Wallis test and the median test if the ANOVA test would prove the data is not normally distributed.