Collection of recipes and some not so well appreciatedd best practices for the impatient..data scientist..
Got missing data just do any one of the following: 1.Remove MISSING rows(sensibly for less than 5% of total data) 2.Impute with some statistic(mean,median etc) 3.Collect more data Almost all of them give biased estimates.
Some thoughts on missingness:(I didn't choose the naming..) Based on the nature of missingness .. people divided three basic regimes: 1.MCAR(Missing Completely At Random): Consider we have a data monster who randomly knocks down some of the observations. 2.MAR(Missing At Random):Instead now he looks at something that exists in the world like he knocks down some sales data on every friday to account for heavy traffic.Randomness but on a deterministic process. 3.MNAR(Missing Not At Random):Your thermometer fails every time at high temperature.You have a perfectly deterministic process.
`