In: Statistics and Probability
People tend to trust numbers and statistics when they are presented with them. As the director of human resources for your city government, you receive a requested analysis of the number of sick days taken by the city staff. It says that in the last six months, men took more time than women, every employee took an average of 3 sick days over the six months, and two employees had gone out on short term disability. You have been asked to recommend a new policy regarding sick days, that may require employees to use vacation days after they use a certain number of sick days.
• Do you trust the data you have to make such a recommendation?
• What else should you know before you make a determination? Think: reliability and validity.
Answer in 200 words
• Do you trust the data you have to make such a recommendation?
First let us analyse the data.
-> In the last six months, men took more time than women.
-> Every employee took an average of 3 sick days over the six months.
-> Two employees had gone out on short term disability.
The trust in data depends on the source. There needs to be nos to show that in the last six months men took more leave than women and the difference should be clear. It can range from 1 to any number. We also know that average is not highy recommended to get inference, other measure such as median should be considered. Median is immune to outliers hence it also caters to the thirs aspect.
• What else should you know before you make a determination?
There are other factors that need to be considered before making a new policy recommendation. Like we need to know the total sample size of the population, to better understand whether our data is skewed or not, i.e, the ratio of men to that of women should be balanced. In case of un-balanced dataset, there is always bias . Moreover, the severity of sick leave should be taken into account and there should be levels. For ecxample if the illness is a mild head-ache it should be assigned a weight of 1 and in case of a surgery or operation it should be assigned a weight of 5. Based on this the number of sick leaves should be mapped. If the staff violates this benchmark then the vacation days should be considered.