, Quantitative research is easy to get wrong
- Statistics is the end-point (first the methodology)
- Before doing statistics, think of alternative explanations & how you can account for those
- If you can’t rule these other explanations out, conclusions must be modest and transparent
o Firms with bad strategy won’t survive
o Mean heart rate by age not many older people were a Fitbit
Simple method is often better than an advanced method
- Strong research designs trump solving problems with advanced methods
- A well-designed experiment can be analyzed with a simple t-test (covid vaccine)
Statistics is the most easy part of quantitative research
- It follows rules and principles
- There is often a right and wrong way of doing something
- It generalizes well across projects
1. Simple linear regression
- Estimate a relation between one x and y variable
o One independent variable input (x)
o One depended variable output (y)
- Scatterplot
- Formula: y(x) = β 0 + β 1 X 1 + ϵ
β 1 : The slope of the line
(i.e., a one unit increase of X 1 results in the β 1increase of y)
β 0 : Intersection y axis
ϵ : Error term distance between every datapoint and the line
- Least squared analysis: Square all the numbers to get all positive numbers
- No error errors are small. (error is to express how good the model is)
- The best possible model is the regression line
- This model might still fit the data poorly
o R-squared: How well fits the model the observed data
- Statistics is the end-point (first the methodology)
- Before doing statistics, think of alternative explanations & how you can account for those
- If you can’t rule these other explanations out, conclusions must be modest and transparent
o Firms with bad strategy won’t survive
o Mean heart rate by age not many older people were a Fitbit
Simple method is often better than an advanced method
- Strong research designs trump solving problems with advanced methods
- A well-designed experiment can be analyzed with a simple t-test (covid vaccine)
Statistics is the most easy part of quantitative research
- It follows rules and principles
- There is often a right and wrong way of doing something
- It generalizes well across projects
1. Simple linear regression
- Estimate a relation between one x and y variable
o One independent variable input (x)
o One depended variable output (y)
- Scatterplot
- Formula: y(x) = β 0 + β 1 X 1 + ϵ
β 1 : The slope of the line
(i.e., a one unit increase of X 1 results in the β 1increase of y)
β 0 : Intersection y axis
ϵ : Error term distance between every datapoint and the line
- Least squared analysis: Square all the numbers to get all positive numbers
- No error errors are small. (error is to express how good the model is)
- The best possible model is the regression line
- This model might still fit the data poorly
o R-squared: How well fits the model the observed data
