NCSS Statistical Software NCSS.

com

Chapter 570

Life-Table Analysis
Introduction
A life table presents the proportion surviving, the cumulative hazard function, and the hazard rates of a large group
of subjects followed over time. The analysis accounts for subjects who die (fail) as well as subjects who are censored
(withdrawn). The life-table method competes with the Kaplan-Meier product-limit method as a technique for
survival analysis. The life-table method was developed first, but the Kaplan-Meier method has been shown to be
superior and with the advent of computers is now the method of choice. However, for large samples, the life-table
method is still popular in that it provides a simple summary of a large set of data.

Construction of a Life Table

We will give a brief introduction to the subject in this section. For a complete account of life-table analysis, we
suggest the books by Lee (1992) and Elandt-Johnson and Johnson (1980). We will use the same terminology as in
the Kaplan-Meier Survival Curves chapter. We suggest that you read the introduction to survival analysis given in
that chapter if you are not familiar with common survival analysis terms such as cumulative survival distribution,
cumulative hazard function, and hazard rates.
A life table is constructed from a set of grouped or ungrouped failure data. The columns of the table are created
using a set of formulas. The rows of the table represent various time intervals. We will now define each of the
columns in the life table. Note, however, that because of the large number of columns required to display all of
the items, there will be several output reports produced.

Time Interval
Each time interval is represented by Tt ≤ T < Tt +1 or [Tt , Tt +1 ) , where t = 1,, s . The interval is from Tt up to but
not including Tt +1 . The intervals are assumed to be fixed. The intervals do not have to be of equal length, but it is
often convenient to make them so.
The midpoint of the interval, Tmt , is defined as half way through the interval.

The width of the interval is bt where bt = Tt +1 − Tt . The width of the last interval, bs , is theoretically infinite, so
items requiring this value will be left blank.

Number Lost to Follow-Up

The number lost to follow-up, lt , is the number of individuals who were loss to observation during this interval, so
their survival status is unknown.

Number Withdrawn Alive

The number withdrawn alive, wt , is the number of individuals who had not died (failed) by the end of the study.

570-1
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Life-Table Analysis

Number Dying
The number dying, d t , is the number of individuals who die (fail) during the interval.

Number Entering the tth Interval

The number entering the tth interval, nt′ , is computed as follows. In the first interval, it is the total sample size. In
the remaining intervals, it is computed using the formula
nt′ = nt′−1 − lt −1 − wt −1 − d t −1

Number Exposed to Risk

The number exposed to risk, nt , is computed using the formula
nt = nt′−1 − 12 (lt −1 + wt −1 )

This formula assumes that times to loss or withdrawal are distributed uniformly across the interval.

Conditional Proportion Dying

The conditional proportion dying, qt , is an estimate of the conditional probability of death in the interval given
exposure to the risk of death in the interval. It is computed using the formula
dt
qt =
nt

Conditional Proportion Surviving

The conditional proportion surviving, pt , is an estimate of the conditional probability of surviving through the
interval. It is computed using the formula
pt = 1 − qt

Cumulative Proportion Surviving

The cumulative proportion surviving, S (Tt ) , is an estimate of cumulative survival rate at time Tt . It is computed
using the formula
S (Tt ) = S (Tt −1 ) pt −1
where
S (T1 ) = 1
The variance of this estimate is itself estimated using the formula
t −1
qj
V [S (Tt )] = S (Tt )
2
∑n
j =1 j pj

Using these estimates, pointwise confidence intervals are given using the Kaplan-Meier product-limit formulas
given in the Kaplan-Meier chapter.

Estimated Death Density Function

The estimated death density function, f (Tmt ) , is an estimate of the probability of dying in this interval per unit
width. At the interval midpoint, it is computed using the formula

570-2
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Life-Table Analysis

S (Tt ) − S (Tt +1 )
f (Tmt ) =
bt
S (Tt )qt
=
bt
The variance of this estimate is itself estimated using the formula

S (Tt ) qt2 t −1  qj pj 
2
V [ f (Tmt )] =
bt
∑  n pj
+ 
n j q j 
j =1  j

Hazard Rate Function

The estimated hazard rate function, h (Tmt ) , is an estimate of the number of deaths per unit time divided by the
average number of survivors at the interval midpoint. It is computed using the formula
f (Tmt )
h (Tmt ) =
S (Tmt )
dt
=
bt (nt − 12 d t )
2qt
=
bt (1 + pt )
The variance of this estimate is itself estimated using the formula

h (Tmt )   h (Tmt )bt  

2 2

V [h (Tmt )] = 1−
nt qt   2 t  
 
Using these estimates, pointwise confidence intervals are given using the cumulative hazard confidence interval
formulas given in the Kaplan-Meier chapter.

Cumulative Hazard Function

The cumulative hazard function, H (Tt ) , is estimated using the Nelson-Aalen method. It is computed using the
formula
t
dj
∑n
~
H (Tt ) =
j =1 j

The variance of this estimate is itself estimated using the formula

[ ] ∑ dn
t
~
V H (Tt ) =
j
2
j =1 j

Using these estimates, pointwise confidence intervals are given using the Nelson-Aalen formulas given in the
Kaplan-Meier chapter.

Median Remaining Lifetime

The median remaining lifetime, MRTt , is the time value at which exactly one-half of those who survived until Tt
are still alive.
To compute this value, find the value j such that S (T j ) ≥ 12 S (Tt ) and S (T j +1 ) < 12 S (Tt ) . Next, computed the median
remaining lifetime using the formula

570-3
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Life-Table Analysis

b j (S (T j ) − 12 S (Tt ))
MRTt = (T j − Tt ) +
S (T j ) − S (T j +1 )

The variance of this estimate is itself estimated using the formula

S (Tt )
2
V (MRTt ) =
[
4ni f (Tmj )
2
]
Using these estimates, pointwise confidence intervals are given using the Nelson-Aalen formulas given in the
Kaplan-Meier chapter. Note that in this case, the confidence intervals are very crude since the MRTt are not
necessarily distributed normally, even in large samples.

Data Structure
Survival datasets require the ending survival time and an indicator of whether an observation was censored or
failed. Additionally, you may also include a frequency variable the gives the count for each row.
The table below shows a dataset from which Lee (1992) constructs a life table. The survival experience of 2418
males with angina pectoris is recorded in years. The life table will use 16 intervals of one year each. These data
are contained in the Lee91 dataset. Note that two rows are required for each data value, one for the failed
individuals and another for the censored individuals.
Lee91 dataset
Time Censor Count
0.5 1 456
1.5 1 226
2.5 1 152
3.5 1 171
4.5 1 135
5.5 1 125
6.5 1 83
7.5 1 74
8.5 1 51
9.5 1 42
10.5 1 43
11.5 1 34
12.5 1 18
13.5 1 9
14.5 1 6
15.5 1 0
0.5 0 0
1.5 0 39
2.5 0 22
3.5 0 23
4.5 0 24
5.5 0 107
6.5 0 133
7.5 0 102
8.5 0 68
9.5 0 64
10.5 0 45
11.5 0 53
12.5 0 33

570-4
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Life-Table Analysis

Lee91 dataset (cont’d)

Time Censor Count
13.5 0 27
14.5 0 23
15.5 0 30

Procedure Options
This section describes the options available in this procedure.

Variables Tab
This panel specifies the variables used in the analysis.

Time Variables
Time Variable
This variable contains the length of time that an individual was observed. This may represent a failure time or a
censor time. Whether the subject actually died is specified by the Censor Variable. Since the values are elapsed
times, they must be positive. Zeroes and negative values are treated as missing values. If you have date values,
you much subtract them so that you have a column of elapsed times.
Time Interval Boundaries
Specify a list of times to be used as boundary points along the time scale. These become the left boundaries of the
time intervals. Care should be taken to specify a left-most boundary that is less than the smallest time value. This
number is often zero.
It is often convenient to make all intervals of the same width, but it is not necessary to do so.
Each interval is closed on the left and open on the right. That is, the interval is T(i) <= T < T(i+1).
Numbers representing the times are separated by blanks or commas. Specify sequences with a colon, putting the
increment inside parentheses. For example: 5:25(5) means 5 10 15 20 25. Avoid negative numbers.
Use ‘(10)’ alone to specify ten, equal-spaced values between zero and the maximum.

Frequency Variable
Frequency Variable
This variable gives the count, or frequency, of the time displayed on that row. This is the number of subjects
represented by each row. When omitted, each row receives a frequency of one. Frequency values should be
positive integers.

Censor Variable
Censor Variable
The values in this variable indicate whether the value of the Time Variable represents a censored time or a failure
time. These values may be text or numeric. The interpretation of these codes is specified by the Failed and
Censored options to the right of this option.
Only two values are used, the Failure code and the Censor code. The Unknown Type option specifies what is to
be done with values that do not match either the Failure code or the Censor code.
Rows with missing values (blanks) in this variable are omitted.

570-5
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Life-Table Analysis

Failed
This value identifies those values of the Censor Variable that indicate that the Time Variable gives a failure time.
The value may be a number or a letter.
We suggest the letter ‘F’ or the number ‘1’ when you are in doubt as to what to use.
Censored
This value identifies those values of the Censor Variable that indicate that the individual recorded on this row was
censored. That is, the actual failure time occurs sometime after the value of the Time Variable.
We suggest the letter ‘C’ or the number ‘0’ when you are in doubt as to what to use.
A censored observation is one in which the time until the event of interest is not known because the individual
withdrew from the study, the study ended before the individual failed, or for some similar reason.
Unknown Censor
This option specifies what the program is to assume about rows whose censor value is not equal to either the
Failed code or the Censored code. Note that observations with missing censor values are always treated as
missing.

• Censored
Observations with unknown censor values are assumed to have been censored.

• Failed
Observations with unknown censor values are assumed to have failed.

• Missing
Observations with unknown censor values are assumed to be missing and they are removed from the analysis.

Group Variable
Group Variable
An optional categorical (grouping) variable may be specified. If it is used, a separate analysis is conducted for
each unique value of this variable.

Options
Confidence Limits
This option specifies the method used to estimate the confidence limits of the survival and hazard values that are
displayed. The options are:

• Linear
This is the classical method which uses Greenwood’s estimate of the variance.

• Log Transform
This method uses the logarithmic transformation of Greenwood’s variance estimate. It produces better limits
than the Linear method and has better small sample properties.

• ArcSine
This method uses the arcsine square-root transformation of Greenwood’s variance estimate to produce better
limits.

570-6
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Life-Table Analysis

Variance
The option specifies which estimator of the variance of the Nelson-Aalen cumulative hazard estimate is to be
used. Three estimators have been proposed. When there are no event-time ties, all three give about the same
results.
We recommend that you use the Simple estimator unless ties occur naturally in the theoretical event times.

• Simple
This estimator should be used when event-time ties are caused by rounding and lack of measurement
precision. This estimate gives the largest value and hence the widest, most conservation, confidence intervals.

• Plug In
This estimator should be used when event-time ties are caused by rounding and lack of measurement
precision.

• Binomial
This estimator should be used when ties occur in the theoretical distribution of event times.

Reports Tab
The following options control which reports are displayed and the format of those reports.

Select Reports
Data Summary - Median Remaining Lifetime
These options indicate whether to display the corresponding report.
Alpha Level
This is the value of alpha used in the calculation of confidence limits. For example, if you specify 0.04 here, then
96% confidence limits will be calculated.

Report Options
Precision
Specify the precision of numbers in the report. A single-precision number will show seven-place accuracy, while
a double-precision number will show thirteen-place accuracy. Note that the reports are formatted for single
precision. If you select double precision, some numbers may run into others. Also note that all calculations are
performed in double precision regardless of which option you select here. Single precision is for reporting
purposes only.
Variable Names
This option lets you select whether to display only variable names, variable labels, or both.
Value Labels
This option lets you select whether to display only values, only value labels, or both for values of the group
variable. Use this option if you want to automatically attach labels to the values of the group variable (like
1=Male, 2=Female, etc.). See the section on specifying Value Labels elsewhere in this manual.

570-7
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Life-Table Analysis

Report Options – Decimal Places

Time
This option specifies the number of decimal places shown on reported time values.
Probability
This option specifies the number of decimal places shown on reported probability and hazard values.
N
This option specifies the number of decimal places shown on the number exposed.

Plots Tab
The following options control the plots that are displayed.

Select Plots
Survival/Reliability Plot - Hazard Rate Plot
These options specify which plots type of plots are displayed. Check the plots that you want to see.

Plot Options – Plot Contents

These options control objects that are displayed on all plots. Click the plot format button to change the plot
settings.
Function Line
Indicate whether to display the estimated survival (Kaplan-Meier) or hazard function on the plots.
C.L. Lines
Indicate whether to display the confidence limits of the estimated function on the plots.

Select Plots – Plots Displayed

Individual-Group Plots
When checked, this option specifies that a separate chart of each designated type is displayed.
Combined Plots
When checked, this option specifies that a chart combining all groups is to be displayed.

Plot Options – Plot Arrangement

These options control the size and arrangement of the plots.
Two Plots Per Line
When a man charts are specified, checking this option will cause the size of the charts to be reduced so that they
can be displayed two per line. This will reduce the overall size of the output.

570-8
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Life-Table Analysis

Storage Tab
These options let you specify if, and where on the dataset, various statistics are stored.
Warning: Any data already in these columns are replaced by the new data. Be careful not to specify columns that
contain important data.

Data Storage Options

Storage Option
This option controls whether the values indicated below are stored on the dataset when the procedure is run.

• Do not store data

No data are stored even if they are checked.

• Store in empty columns only

The values are stored in empty columns only. Columns containing data are not used for data storage, so no
data can be lost.

• Store in designated columns

Beginning at the Store First Item In column, the values are stored in this column and those to the right. If a
column contains data, the data are replaced by the storage values. Care must be used with this option because
it cannot be undone.
Store First Item In
The first item is stored in this column. Each additional item that is checked is stored in the columns immediately
to the right of this column.
Leave this value blank if you want the data storage to begin in the first blank column on the right-hand side of the
data.
Warning: any existing data in these columns is automatically replaced, so be careful.

Data Storage Options – Select Items to Store on

the Spreadsheet
Group - Median R.L. UCL
Indicate whether to store these values, beginning at the column indicated by the Store First Item In option.

570-9
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Life-Table Analysis

Example 1 – Creating a Life Table

This section presents an example of how to create a life table. This example will use the survival data contained in
the Lee91 dataset .
You may follow along here by making the appropriate entries or load the completed template Example 1 by
clicking on Open Example Template from the File menu of the Life-Table Analysis window.

1 Open the Lee91 dataset.

• From the File menu of the NCSS Data window, select Open Example Data.
• Click on the file Lee91.NCSS.
• Click Open.

2 Open the Life-Table Analysis window.

• Using the Analysis menu or the Procedure Navigator, find and select the Life-Table Analysis procedure.
• On the menus, select File, then New Template. This will fill the procedure with the default template.

3 Specify the variables.

• On the Life-Table Analysis window, select the Variables tab.
• Set the Time Variable to Time.
• Set the Time Interval Boundaries to 0:15(1).
• Set the Frequency Variable to Count.
• Set the Censor Variable to Censor.

4 Specify the plots.

• On the Life-Table Analysis window, select the Plots tab.
• Check the Hazard Function Plot box.
• Check the Hazard Rate Plot box.

5 Run the procedure.

• From the Run menu, select Run Procedure. Alternatively, just click the green Run button.

Data Summary Section

Type of
Observation Rows Count Minimum Maximum
Failed 15 1625 0.5 14.5
Censored 15 793 1.5 15.5
Total 30 2418 0.5 15.5

This report displays a summary of the amount of data that were analyzed. Scan this report to determine if there
were any obvious data errors by double checking the counts and the minimum and maximum times.

570-10
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Life-Table Analysis

Life-Table Analysis Detail Section

No. No. Conditonal Cumulative
Start No. No. Exp'd Proportion Proportion Hazard
Time Int'l Lost Died to Risk Surviving Surviving Rate
0.0 2418 0 456 2418.0 0.81141 1.00000 0.20822
1.0 1962 39 226 1942.5 0.88366 0.81141 0.12353
2.0 1697 22 152 1686.0 0.90985 0.71701 0.09441
3.0 1523 23 171 1511.5 0.88687 0.65237 0.11992
4.0 1329 24 135 1317.0 0.89749 0.57856 0.10804
5.0 1170 107 125 1116.5 0.88804 0.51926 0.11860
6.0 938 133 83 871.5 0.90476 0.46112 0.10000
7.0 722 102 74 671.0 0.88972 0.41721 0.11672
8.0 546 68 51 512.0 0.90039 0.37120 0.10483
9.0 427 64 42 395.0 0.89367 0.33422 0.11230
10.0 321 45 43 298.5 0.85595 0.29868 0.15523
11.0 233 53 34 206.5 0.83535 0.25566 0.17942
12.0 146 33 18 129.5 0.86100 0.21356 0.14938
13.0 95 27 9 81.5 0.88957 0.18388 0.11688
14.0 59 23 6 47.5 0.87368 0.16357 0.13483
15.0 30 30 0 15.0 1.00000 0.14291

This report displays the standard life table. The formulas used were presented earlier.
Time
This is the left boundary of the time interval reported on this line. The right boundary is the entry on the following
line. Each interval is represented by Tt ≤ T < Tt +1 .

No. Start Int’l

This is the number entering the tth interval. In the first interval, it is the total sample size.
No. Lost
This is the number lost to follow-up and the number withdrawn from the study alive.
No. Died
This is the number of individuals who died (failed) during the interval.
No. Exp’d to Risk
This is the average number exposed to risk in the interval. It is calculated under the assumption that losses and
withdrawals are distributed uniformly across the interval.
Conditional Proportion Surviving
This is the conditional proportion surviving through the interval.
Cumulative Proportion Surviving
This is the estimate of the survivorship function, S (Tt ) . It is also called the cumulative survival rate at time Tt . It
is the probability of surviving to the start of the interval and then through the interval.
Hazard Rate
This is the estimated hazard rate function, h(Tmt ) . It is an estimate of the number of deaths per unit time divided by
the average number of survivors computed at the interval midpoint.

570-11
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Life-Table Analysis

Life-Table Analysis Summary Section

Cumulative Cumulative Death Median No.
Proportion Hazard Hazard Density Remaining Starting
Time Surviving Function Rate Function Lifetime Int'l
0.0 1.00000 0.18859 0.20822 0.18859 5.3 2418
1.0 0.81141 0.30493 0.12353 0.09440 6.2 1962
2.0 0.71701 0.39508 0.09441 0.06464 6.3 1697
3.0 0.65237 0.50822 0.11992 0.07380 6.2 1523
4.0 0.57856 0.61072 0.10804 0.05931 6.2 1329
5.0 0.51926 0.72268 0.11860 0.05813 5.9 1170
6.0 0.46112 0.81792 0.10000 0.04392 5.6 938
7.0 0.41721 0.92820 0.11672 0.04601 5.2 722
8.0 0.37120 1.02781 0.10483 0.03697 4.9 546
9.0 0.33422 1.13414 0.11230 0.03554 4.8 427
10.0 0.29868 1.27819 0.15523 0.04303 4.7 321
11.0 0.25566 1.44284 0.17942 0.04209 233
12.0 0.21356 1.58184 0.14938 0.02968 146
13.0 0.18388 1.69227 0.11688 0.02031 95
14.0 0.16357 1.81858 0.13483 0.02066 59
15.0 0.14291 1.81858 30

This report displays the most interesting quantities from a life table. The formulas used were presented earlier.
Time
This is the left boundary of the time interval reported on this line. The right boundary is the entry on the following
line. Each interval is represented by Tt ≤ T < Tt +1 .

Cumulative Survival
This is the estimate of the survivorship function, S (Tt ) . It is also called the cumulative survival rate at time Tt . It
is the probability of surviving to the start of the interval and then through the interval.
Cumulative Hazard Function
This is the estimate of the cumulative hazard function, H (Tt ) .

Hazard Rate
This is the estimated hazard rate function, h(Tmt ) . It is an estimate of the number of deaths per unit time divided by
the average number of survivors computed at the interval midpoint.
Death Density Function
This is the estimated death density function, f (Tmt ) . It is an estimate of the probability of dying at the interval
midpoint.
Median Remaining Lifetime
This is the median remaining lifetime, MRTt . It is the time value at which exactly one-half of those who survived
until Tt are still alive.

No. Start Int’l

This is the number entering the tth interval. In the first interval, it is the total sample size.

570-12
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Life-Table Analysis

Survival Analysis Section

Cumulative Standard Lower Upper
Time Survival Error 95% C.L. 95% C.L.
0.0 1.00000 0.00000 1.00000 1.00000
1.0 0.81141 0.00796 0.79582 0.82701
2.0 0.71701 0.00918 0.69902 0.73500
3.0 0.65237 0.00973 0.63329 0.67145
4.0 0.57856 0.01014 0.55869 0.59844
5.0 0.51926 0.01030 0.49906 0.53945
6.0 0.46112 0.01038 0.44078 0.48147
7.0 0.41721 0.01045 0.39672 0.43769
8.0 0.37120 0.01058 0.35046 0.39193
9.0 0.33422 0.01072 0.31322 0.35523
10.0 0.29868 0.01089 0.27734 0.32003
11.0 0.25566 0.01112 0.23385 0.27746
12.0 0.21356 0.01140 0.19123 0.23590
13.0 0.18388 0.01177 0.16082 0.20694
14.0 0.16357 0.01226 0.13954 0.18760
15.0 0.14291 0.01330 0.11684 0.16898

This report displays the life-table survival distribution along with confidence limits. The formulas used were
presented earlier.
Time
This is the left boundary of the time interval reported on this line. The right boundary is the entry on the following
line. Each interval is represented by Tt ≤ T < Tt +1 .

Cumulative Survival
This is the estimate of the survivorship function, S (Tt ) . It is also called the cumulative survival rate at time Tt . It
is the probability of surviving to the start of the interval and then through the interval.
Standard Error
This is the large-sample estimate of standard error of the survival function. It is a measure of the precision of the
survival estimate.
Lower and Upper Confidence Limits
The lower and upper confidence limits provide a pointwise confidence interval for the survival function. These
limit are constructed so that the probability that the true survival probability lies between them is 1 − α . Note that
these limits are constructed for a single time point. Several of them cannot be used together to form a confidence
band such that the entire survival function lies within the band.
Three difference confidence intervals are available. All three confidence intervals perform about the same in large
samples. The linear (Greenwood) interval is the most commonly used. However, the log-transformed and the
arcsine-square intervals behave better in small to moderate samples, so they are recommended. The formulas for
these limits were given at the beginning of the chapter and are not repeated here.

570-13
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Life-Table Analysis

Cumulative Hazard Section

Cumulative Standard Lower Upper
Time Hazard Error 95% C.L. 95% C.L.
0.0 0.18859 0.00883 0.17128 0.20589
1.0 0.30493 0.01174 0.28192 0.32795
2.0 0.39508 0.01383 0.36797 0.42220
3.0 0.50822 0.01632 0.47624 0.54020
4.0 0.61072 0.01855 0.57437 0.64708
5.0 0.72268 0.02108 0.68137 0.76399
6.0 0.81792 0.02353 0.77180 0.86403
7.0 0.92820 0.02679 0.87568 0.98072
8.0 1.02781 0.03021 0.96860 1.08702
9.0 1.13414 0.03438 1.06676 1.20151
10.0 1.27819 0.04080 1.19824 1.35815
11.0 1.44284 0.04961 1.34560 1.54009
12.0 1.58184 0.05946 1.46531 1.69837
13.0 1.69227 0.06993 1.55521 1.82932
14.0 1.81858 0.08689 1.64829 1.98888
15.0 1.81858 0.08689 1.64829 1.98888

This report displays estimates of the cumulative hazard function. The formulas used were presented earlier.
Time
This is the left boundary of the time interval reported on this line. The right boundary is the entry on the following
line. Each interval is represented by Tt ≤ T < Tt +1 .

Cumulative Hazard
This is the Nelson-Aalen estimate of the cumulative hazard function, H (Tt ) .

Standard Error
This is the estimated standard error of the above cumulative hazard function. The formula used was specified
under the Variables tab in the Variance box. The standard error is the square root of this variance.
Lower and Upper Confidence Limits
The lower and upper confidence limits provide a pointwise confidence interval for the cumulative hazard at each
time point. These limits are constructed so that the probability that the true cumulative hazard lies between them
is 1 − α . Note that these limits are constructed for a single time point. Several of them cannot be used together to
form a confidence band such that the entire cumulative hazard function lies within the band.
Three difference confidence intervals are available. All three confidence intervals perform about the same in large
samples. The linear (Greenwood) interval is the most commonly used. However, the log-transformed and the
arcsine-square intervals behave better in small to moderate samples, so they are recommended. The formulas for
these limits were given at the beginning of the chapter and are not repeated here.

570-14
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Life-Table Analysis

Hazard Rate Section

Hazard Standard Lower Upper
Time Rate Error 95% C.L. 95% C.L.
0.0 0.20822 0.00970 0.18921 0.22723
1.0 0.12353 0.00820 0.10746 0.13961
2.0 0.09441 0.00765 0.07942 0.10940
3.0 0.11992 0.00915 0.10197 0.13786
4.0 0.10804 0.00929 0.08984 0.12624
5.0 0.11860 0.01059 0.09784 0.13935
6.0 0.10000 0.01096 0.07851 0.12149
7.0 0.11672 0.01355 0.09017 0.14327
8.0 0.10483 0.01466 0.07610 0.13356
9.0 0.11230 0.01730 0.07839 0.14621
10.0 0.15523 0.02360 0.10898 0.20149
11.0 0.17942 0.03065 0.11935 0.23948
12.0 0.14938 0.03511 0.08056 0.21819
13.0 0.11688 0.03889 0.04065 0.19311
14.0 0.13483 0.05492 0.02719 0.24247
15.0

This report displays estimates of the hazard rates at the midpoints of each of the time intervals. The formulas used
were presented earlier.
Time
This is the left boundary of the time interval reported on this line. The right boundary is the entry on the following
line. Each interval is represented by Tt ≤ T < Tt +1 . Note that the hazard rate is actually computed at the midpoint of
each interval.
Cumulative Hazard
This is the estimate of the hazard rate, h(Tmt ) .

Standard Error
This is the estimated standard error of the above hazard rate. The formula used was given earlier. The standard
error is the square root of this variance.
Lower and Upper Confidence Limits
The lower and upper confidence limits provide a pointwise confidence interval for the hazard rate at the midpoint
of the time interval. These limits are constructed so that the probability that the true hazard rate lies between them
is 1 − α . Note that these limits are constructed for a single time point. Several of them cannot be used together to
form a confidence band such that the entire hazard rate lies within the band.
Three difference confidence intervals are available. All three confidence intervals perform about the same in large
samples. The linear (Greenwood) interval is the most commonly used. However, the log-transformed and the
arcsine-square intervals behave better in small to moderate samples, so they are recommended. The formulas for
these limits were given at the beginning of the chapter and are not repeated here.

570-15
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Life-Table Analysis

Death Density Function Section

Death Standard Lower Upper
Time Density Error 95% C.L. 95% C.L.
0.0 0.18859 0.00796 0.17299 0.20418
1.0 0.09440 0.00598 0.08269 0.10612
2.0 0.06464 0.00507 0.05471 0.07458
3.0 0.07380 0.00543 0.06317 0.08444
4.0 0.05931 0.00495 0.04961 0.06900
5.0 0.05813 0.00503 0.04827 0.06800
6.0 0.04392 0.00469 0.03472 0.05311
7.0 0.04601 0.00518 0.03587 0.05615
8.0 0.03697 0.00502 0.02713 0.04682
9.0 0.03554 0.00531 0.02513 0.04594
10.0 0.04303 0.00627 0.03074 0.05532
11.0 0.04209 0.00685 0.02867 0.05551
12.0 0.02968 0.00668 0.01659 0.04278
13.0 0.02031 0.00651 0.00754 0.03307
14.0 0.02066 0.00804 0.00491 0.03641
15.0

This report displays estimates of the hazard rates at the midpoints of each of the time intervals. The formulas used
were presented earlier.
Time
This is the left boundary of the time interval reported on this line. The right boundary is the entry on the following
line. Each interval is represented by Tt ≤ T < Tt +1 . Note that the hazard rate is actually computed at the midpoint of
each interval.
Death Density
This is the estimate of the death density, f (Tmt ) .

Standard Error
This is the estimated standard error of the above density. The formula used was given earlier. The standard error
is the square root of this variance.
Lower and Upper Confidence Limits
The lower and upper confidence limits provide a pointwise confidence interval for the death density at the
midpoint of the time interval. These limits are constructed so that the probability that the true density lies between
them is 1 − α .
Three difference confidence intervals are available. All three confidence intervals perform about the same in large
samples. The linear (Greenwood) interval is the most commonly used. However, the log-transformed and the
arcsine-square intervals behave better in small to moderate samples, so they are recommended. The formulas for
these limits were given at the beginning of the chapter and are not repeated here.

570-16
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Life-Table Analysis

Median Remaining Lifetime Section

Median
Remaining Standard Lower Upper
Time Lifetime Error 95% C.L. 95% C.L.
0.0 5.3 0.17491 5.0 5.7
1.0 6.2 0.20006 5.9 6.6
2.0 6.3 0.23614 5.9 6.8
3.0 6.2 0.23609 5.8 6.7
4.0 6.2 0.18526 5.9 6.6
5.0 5.9 0.18059 5.6 6.3
6.0 5.6 0.18554 5.2 6.0
7.0 5.2 0.27129 4.6 5.7
8.0 4.9 0.27632 4.4 5.5
9.0 4.8 0.41408 4.0 5.6
10.0 4.7 0.41835 3.9 5.5
11.0
12.0
13.0
14.0
15.0

This report displays estimates of the median remaining lifetime of those who are alive at the beginning of the
interval. The formulas used were presented earlier.
Time
This is the left boundary of the time interval reported on this line. The right boundary is the entry on the following
line. Each interval is represented by Tt ≤ T < Tt +1 .

Median Remaining Lifetime

This is the estimate of the median remaining lifetime of an individual who survives to the beginning of this
interval.
Standard Error
This is the estimated standard error of the above lifetime. The formula used was given earlier.
Lower and Upper Confidence Limits
The lower and upper confidence limits provide a pointwise confidence interval for the hazard rate at the midpoint
of the time interval. These limits are constructed so that the probability that the true remaining lifetime lies
between them is 1 − α .
These confidence intervals are highly approximate. They depend on the assumption that the median remaining
lifetime is normally distributed. This may not be true even in large samples.

570-17
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Life-Table Analysis

Survival Plot

This plot shows the survivorship function. If there are several groups, a separate line is drawn for each group.

Hazard Function Plot

This plot shows the Nelson-Aalen cumulative hazard function. If you have several groups, then a separate line is
drawn for each group.

570-18
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Life-Table Analysis

Hazard Rate Plot

This plot shows the hazard rates. Note the unusual step-like appearance of the plot because the hazard rates are
assumed constant for the duration of the interval.

Validation of Life-Table Estimator using Lee (1992)

This section presents validation of our life-table estimator. Lee (1992) presents an example on page 91 of a
calculated life table. We will include the results of one line of that table so that you can compare those results
with those produced by this program. If you compare these values with those shown above, you can validate that
NCSS provides the correct results.

Parameter Value
Time 3
Lost 23
Dying 171
Entering 1523
Exposed 1511.5
Proportion Dying 0.1131
Proportion Surviving 0.8869
S(T) 0.6524
S.E. S(T) 0.0097
h(T) 0.1199
S.E. h(T) 0.0092
Median Rem. Lifetime 6.23
S.E. MRL(T) 0.9
f(T) 0.0738
S.E. f(T) 0.0054

570-19
© NCSS, LLC. All Rights Reserved.