DISCUSS SAMPLE DISTRIBUTIONS

DISCUSS SAMPLE DISTRIBUTIONS

The sample size is 580, which is quite a large sample. As this sample size is quite large it appears to be a good representation of the whole population. This data was collected from a sample of VLBW babies born during the 1980's in the USA.

The sample of 580 birth weights of VLBW babies represents how I would have predicted the whole population of birth weights of VLBW babies to look and so I am comfortable to proceed with the analysis based on this sample.

Looking at the sample of 581 VLBW baby weights, we see an approximately normal distribution, with a little bit of a left skew.

The median birth weight of VLBW babies that do not survive is 850g which is 350g lighter than the median birth weight of VLBW babies that do survive (which is 1200g). As the sample size is 581 which is a good representation of the population of VLBW babies born in America in the 1980s it appears my theory that VLBW babies tend to be more likely to survive when they are heavier is true.

For this sample we can see from the graph that the median birth-weight for infants that survived is quite a bit higher (170g higher) than the upper quartile of birth-weight of VLBW babies which is approximately 1030g The upper quartile of VLBW babies that survived is substantially higher than the whole middle 50% of VLBW babies that didn't survive. The upper quartile of VLBW babies that survived lies at 1350g which is shifted further up the scale by 320g than the middle 50% of VLBW babies that died. This again enforces my belief that the lower the birth weight of VLBW babies, the higher the risk of death.

For this sample the distribution for the VLBW babies that did survive is skewed to the left, whilst the distribution for the VLBW babies that did not survive is only partially skewed to the right. This appears to be because the data for the VLBW babies that did not survive is very evenly spread whilst the data for the VLBW babies that did survive seems to take a bell shape with the majority of the data between 1000-1500g. This is relatively close to what I had expected the shape of the data to look like and again seems to prove that VLBW with lower birth weights are less likely to survive that VLBW babies with higher birth weights,

In this sample there is a lot less data provided for the VLBW babies that did not survive as opposed to the amount of data provided for the VLBW babies that did survive. It would be beneficial to take a different sample with a more even amount of data each way so we can fairly compare the birth weights of VLBW babies that did survive and VLBW babies that did not survive.

STEP 4: DISCUSS SAMPLING VARIABILITY, INCLUDING VARIABILITY OF ESTIMATES

Based on this sample estimate it appears that the difference between the median birth weight of VLBW babies that survive and VLBW babies that do not survive is 350g. However it must be recognized that thing conclusion is drawn from just one sample and could be due to chance alone. In order to refine this, multiple samples would need to be taken and analysed.

This problem is called sampling variability and is an issue commonly faced when using a sample statistic to estimate a population parameter.