Normal Probabilities on the TI-83
Tom Linton, http://www.central.edu/homepages/lintont/
This file is a brief summary of how to utilize the TI-83 calculator
to:
-
Calculate probabilities associated with intervals
of values from a normal distribution;
-
Draw plots of the "area under a normal density function";
and
-
Calculate inverse values (from a normal distribution)
which correspond to a given percentage, such as "what value has 60% of
the data below it?"
The three main commands we'll use are all located on the distribution
menu (or its DRAW sub-menu). The distribution menu is the [2nd]
function of the [VARS] key. It is located near the top right corner
of the keypad and is labeled [DISTR].
Calculating Normal Probabilities
Two sided intervals
The normalcdf command, [2nd] [DISTR] [2], makes
calculating probabilities for intervals of normal values straightforward.
You must tell this command where the interval begins (i.e. give the left
or lower endpoint of the interval), where the interval ends (its right
or upper endpoint) and optionally both the mean and standard deviation
(in that order) of the distribution. If you leave off the mean and standard
deviation parameters, the TI-83 assumes the mean is zero and the standard
deviation is 1 (i.e. the standard normal distribution is used). The general
syntax of the normalcdf command is:
normalcdf( start, end, mean, std_dev)
The TI-83 will return the percentage (as a decimal number from 0 to 1)
of data points x in a normal distribution with the given mean and standard
deviation which lie between x = start and x = end. For example,
if the heights of 6 year old girls are normally distributed with a mean
of 46 inches and a standard deviation of 2.7 inches, then we can calculate
the likelihood that a randomly chosen 6 year old girl has height between
44 and 48 inches with the command
normalcdf(44,48,46,2.7)
The TI-83 returns 0.54115, indicating that about 54% of six year old girls
have heights between 44 and 48 inches. Use your calculator to verify this
command by entering the centered command above, pressing [ENTER] and checking
that your calculator returns the answer .54115 (roughly).
One sided intervals
Many questions dealing with normally distributed values have only one endpoint.
For example, what percentage of six year girls have a height less than
44 inches? or greater than 42 inches? These questions involve plus or minus
infinity as the left or right endpoint of the interval, and are called
one sided intervals. If you are looking for the likelihood that a six year
old girl has a height which is less than 44 inches, you are really looking
for the percentage of six year old girls with a height between minus infinity
inches and 44 inches. There are several ways to deal with one sided intervals
like this (where either the start is minus infinity, or the end is positive
infinity). One such method relies on the fact that there is virtually no
probability associated to normal values which are more than 5 standard
deviations below, or 5 standard deviations above the mean. Using this fact,
we can freely replace minus infinity with any value which is 5 (or more)
standard deviations below the mean, and instead of using plus infinity,
simply use a number which is 5 (or more) standard deviations above the
mean. Here is an example using the normal distribution of the heights of
6 year old girls ( i.e. normal with a mean of 46 inches and a std. dev.
of 2.7 inches).
A certain water-slide park requires that all sliders must be 48 inches
or taller to go down their water slides. What percentage of 6 year old
girls are allowed to slide in this park? The correct answer would correspond
to the area under the normal density function (the one with a mean of 46
and a standard deviation of 2.7) from x = 48 to infinity. We can essentially
calculate this with the command normalcdf(48, 59.5, 46, 2.7) = 0.2294.
Thus, only about 23% of six year old girls are tall enough for this water
park.
NOTE: 59.5 = 46 + 5*2.7 (mean + 5 * std_dev) and you can let
the calculator do this arithmetic. Rather than figuring out what 46 + 5*2.7
is, you can simply enter the "ending value" as
46 + 5*2.7
the calculator will do the arithmetic automatically. In a similar fashion,
the percentage of 6 year old girls whose heights are less than 40 inches
can be accessed by pressing
[2nd][DISTR] [2] [4][6][-][5][*][2][.][7][,] 40, 46, 2.7 )
and then pressing [ENTER]. The result is 0.01313..., so only about 1.3%
of six year old girls are less than 40 inches tall.
Normal areas plots
It is always a good idea to visualize the areas you are calculating. The
ShadeNorm
command (located on the [DRAW] sub-menu of the [DISTR] menu) does just
this. You pass the ShadeNorm command exactly the same values as
you pass to the normalcdf command (start, end, and an optional
mean and std. dev.), but you should set up your plot window before giving
the ShadeNorm command. In general, set X to run from 4 std. dev.'s
below the mean to 4 std. dev.'s above the mean (so Xmin = mean - 4 * std_dev,
Xmax = mean + 4*std_dev). You can run the y values from a small negative
number (or zero) to 0.4 / std_dev. The value 0.4 / std_dev will normally
give a good view in the y direction. Because the ShadeNorm command
actually displays text on your screen, I prefer setting YMin = -YMax, which
usually avoids the text overlapping the picture. Here are the steps I would
use to draw a picture of the percentage of six year old girls with heights
from 44 to 48 inches.
-
Turn off all STAT PLOTS and functions on the Y= screen.
-
From the WINDOW screen, set XMin = 46 - 4*2.7, XMax = 46 + 4*2.7, YMin
= -0.4 / 2.7, and YMax = 0.4 / 2.7.
-
Press [2nd][DISTR][right arrow key] to access the DRAW submenu of the [DISTR]
menu. Press [ENTER] to paste a copy of the ShadeNorm command to
the homescreen. Finally, give the same arguments as you would to the normalcdf
command, namely 44, 48, 46, 2.7) and press [ENTER]. You should
see a plot similar to the one below.
Normally, the drawing displayd will dissappear once you alter any plotting
value. You can also erase the drawing by pressing [2nd][DRAW][ENTER]
which executes the ClrDraw command.
Inverse values
The inverse normal command, InvNorm, allows you to compute the value
of a normal distribution which lies at a certain percentage in the distribution,
for example, which x has 92% of the distribution below it? The first argument
must be a decimal between 0 and 1. It represents the percentile value you
are looking for (0.25 for the first quartile, .99 for the data point with
only 1% of the distribution to its right). You can optionally pass the
InvNorm command the mean and standard deviation of the distribution.
If you leave these values off, a mean of 0 and standard deviation of 1
are used. As an example, the start of the third quartile (75% mark) for
heights of six year old girls is obtained with the command
InvNorm(.75,46,2.7)
by pressing [2nd][DISTR] [3] and then entering the arguments and
closing parenthesis. The result (check for yourself) is 47.82 inches, so
75% of all six year old girls are shorter than 47.82 inches, equivalently,
25% of all six year old girls are 47.82 inches or taller.