Timeseries, Correlation And Regression Flashcards
What are the 4 components in timeseries analysis
Seasonal: a pattern that repeats in similar way each year
Cyclical: varies over longer periods according to business cycles (booms and busts)
Trend: An upward or downward trend over a period of time, possibly after removing seasonal variations.
Random: noise that has no identifiable pattern. Residual after all other patterns have been accounted for.
What is meant when two variables are said to be correlated.
Correlated: as one variable increases, the other variable decreases (negative) or increases (positive correlation). There can be more or less chaotic relationship. Less noise means more correlated. If uncorrelated, the way one variable varies does not affect the outcome for the other variable. One cannot use one to predict the other.
What is regression analysis and how does it affect relationships
In regression, you fit a line to a set of data. If values are more scattered and low correlation, the line will not fit so accurately.
The p-value tells the likelihood that the coefficient could be zero. E.g. the intercept coefficient of 423.1 has a very low p-value, zero, so this says that the intercept is definitely a positive value. C.I is [389 to 457] and is range we’d expect the true value to be.
The slope has a low p-value, 3.2%, and is therefore significant also. The confidence interval, [-0.01 to -0.25], shows where we expect true slope to be. It is definitely negative but could be fairly low slope.
