qid 1172

[aspiringmd1015]

i'm pretty decent at epidemiology/biostats, but this one explanation sort of confused me where it says in the last paragraph: if SE had been given instead of SD, investigators would be 95% confident(now theyre talking about confidence intervals) that the true average folate level in the population lies within the mean +/- 2 SE(which is also the range that would include 95% of sample means calculated from repeated sampling of the same size from the population)

---

Members don't see this ad.

 

[zhopv10]

Hard to tell where the confusion lies from what you gave here. I assume that perhaps there is some confusion between SE and SD of the sample data?

The explanation is talking about the idea that when working with experimental data you are working with a sample of the original population and you are trying to approximate the sample mean (true mean). So a 95% CI is suggesting that you are 95% confident the true mean lies within that range of measurements. So you use (SD of the sample data)/sqrt. So 95%CI= mean+or-1.96SD/sqrt. So this accounts for the ability of the data to estimate the true mean of the population by taking into account the spread and sample size. This can be seen by the fact that as the n on the bottom approaches infinity (or as the SD approaches 0) the SE approaches 0 I.e you are exactly approximating the true (population) mean. So the bigger the sample size (and/or the smaller the scatter), the more narrow the CI, and the more precise the estimate.

So just to back up a little SD just measures the spread of your data and isn't, by itself used for CI calculation (as a side note, as the size of your sample increases this SD should approach the population SD). So using just SD you could say that 95% of your measurements are between 1.96SD's of the mean but again that is spread not CI. SE takes SD into account and teams it with sample size to help estimate the precision of your estimate of the mean. Thought of another way. The SE can be seen if you replicated the experiment a bunch (with the same sample size) and then calculated the SD of those measurements would approximate your calculated SE. As the sample size gets bigger those estimates get closer to approximating the true mean.

Simply put: SD is a measure of the scatter and SE quantifies how precisely you know the measure of the true mean, taking into account the scatter (SD) and the sample size sqrt.

Hopefully that is helpful...if it's not perhaps consider stating the question you had in mind, or where the confusion was stemming from.


Sent from my iPhone using SDN mobile app

 

[zhopv10]

@dempty Just to check with the expert, did I get that right? Lol


Sent from my iPhone using SDN mobile app

 

[dempty]

Hard to tell where the confusion lies from what you gave here. I assume that perhaps there is some confusion between SE and SD of the sample data?

The explanation is talking about the idea that when working with experimental data you are working with a sample of the original population and you are trying to approximate the sample mean (true mean). So a 95% CI is suggesting that you are 95% confident the true mean lies within that range of measurements. So you use (SD of the sample data)/sqrt. So 95%CI= mean+or-1.96SD/sqrt. So this accounts for the ability of the data to estimate the true mean of the population by taking into account the spread and sample size. This can be seen by the fact that as the n on the bottom approaches infinity (or as the SD approaches 0) the SE approaches 0 I.e you are exactly approximating the true (population) mean. So the bigger the sample size (and/or the smaller the scatter), the more narrow the CI, and the more precise the estimate.

So just to back up a little SD just measures the spread of your data and isn't, by itself used for CI calculation (as a side note, as the size of your sample increases this SD should approach the population SD). So using just SD you could say that 95% of your measurements are between 1.96SD's of the mean but again that is spread not CI. SE takes SD into account and teams it with sample size to help estimate the precision of your estimate of the mean. Thought of another way. The SE can be seen if you replicated the experiment a bunch (with the same sample size) and then calculated the SD of those measurements would approximate your calculated SE. As the sample size gets bigger those estimates get closer to approximating the true mean.

Simply put: SD is a measure of the scatter and SE quantifies how precisely you know the measure of the true mean, taking into account the scatter (SD) and the sample size sqrt.

Hopefully that is helpful...if it's not perhaps consider stating the question you had in mind, or where the confusion was stemming from.


Sent from my iPhone using SDN mobile app

@dempty Just to check with the expert, did I get that right? Lol


Sent from my iPhone using SDN mobile app

Somehow, this never showed up in my feed until today ....not sure why that happened, but it's getting a bump . More or less, you've given a good explanation. I would just clarify that the SE and the SD are both measures of variability/spread. The SD is for individual observations and used to describe the sample. The SE is variability (also interpreted as precision) surrounding our point estimate of the population parameter (in other words, the SE helps describe the distribution of sample means at a given sample size, for example-- you nailed it in saying that the SE is the SD for a distribution of sample means [or some other sample statistic]). You use the SD to describe the sample whereas the SE is used to make inferences about the population parameter value.