Outcome 3 - Statistics Flashcards
Why stats?
- it allows us to see if our measurements are sufficiently close to the true rate of decay (since it is random and will never be the same each time)
- help distinguish errors from randomness of decay
What is PRECISION?
How reproducible the results are
What is ACCURACY?
How close the measurement is to the true value
What is RELIABILITY?
The level of reproducibility
What is CONSTANCY?
Reproducibility over time
Consistent measurements day after day
What is SENSITIVITY in DIAGNOSIS?
An operating characteristic that:
- measures the ability of a test to detect a disease/condition when it is truly PRESENT
What is SENSITIVITY in EQUIPMENTS?
- Minimum detection limit of a radiation detector
- ability to detect small changes in radioactivity present
What is SPECIFICITY in DIAGNOSIS?
An operating characteristic that:
- measures the ability of a tad to exclude the presence of a disease/condition when it is truly NOT PRESENT
What are the 5 types of errors?
1/2. Determinate or Systematic
3/4. Indeterminate or Random
5. Blunder (user error)
What is the difference between Determinate/Systematic errors and Indeterminate/Random errors?
Determinate/Systematic - results differ from the true value in a consistent way
Indeterminate/Random - inherent differences in the data that can’t be controlled (e.g. randomness of decay)
What is usually the cause of a determinate/systematic error?
usually an issue with equipment QC/calibration
What are the two types of distributions?
- continuous variables
- discrete variables
What is the difference between continuous and discrete variables?
continuous = variables that have any real number (including decimals)
discrete = variables that are limited to whole numbers
What distributions are used for continuous variables? discrete?
continuous - gaussian distribution
discrete - poisson distribution
What is Gaussian statistics?
- used for LARGE sample sizes
- allows the assumption of the sample to represent a “true” state of the entire population
- uses normal distribution or bell curves
What does population mean?
all possible values for a particular characteristic
What does the Y axis in a distribution graph represent?
Y axis = the frequency or probability density
What does the X axis in a distribution graph represent?
X axis = the value of the variable being assessed
What are central tendencies?
mean, mode, median
How do you measure dispersion?
range, variance, standard deviation
How do you calculate range?
difference between highest and lowest value
How do you calculate variance?
sum of the (differences of all values from the mean) squared then divided by sample size - 1
What is the measurement of dispersion?
it is how much the values in the sample vary from the center
What is standard deviation?
the square root of the variance
What is the co-efficient of variation (cv or v)? Why is it important?
- standard deviation as a percentage of the mean
- allows for the comparison between population with different means
Large variance (s^2) = ?
Small variance (s^2) = ?
large = wide divergence from average
small = narrow divergence from average
What are confidence intervals?
pattern that let’s you make assumptions on how many counts/measurements will fall within which interval
difference between true mean and observed mean?
true = population mean
observed = sample mean
(sample - small group within a population; population - all possible values for a particular characteristic)
In a normal distribution, what is the percentage of data that falls within +/- 1SD? +/- 2SD? +/- 3SD?
+/- 1SD = 68.3% (68%)
+/- 2SD = 95.5% (96%)
+/- 3SD = 99.7% (99%)
What characteristics are required to apply Poisson distribution?
- whole numbers greater than 0
- events measured aren’t related to each other
- probability of each event is the same
- events are random, but there is a known rate of occurrence
What is the difference of SD in Gaussian vs. Poisson?
Gaussian - SD is INDEPENDENT of mean
Poisson - SD is DEPENDENT on mean
What does the SD of a single measurement indicate?
It is the prediction of how much variation there would be if there were repeated measurements
of counts ____ = cv ______
Increases, decreases
What is the minimum level of confidence in NM?
2% error with a 95% confidence level (3SD)
Therefore, a minimum of 10000 counts for statistical validity
What is a Levy-Jennings graph used for?
To see the performance of the detector over time.
What is a chi-squared test used for?
Sees if the detector demonstrates the predicted level of variation between measurements
