SIMSTAT: Summary
Using SIMSTAT for statistical analysis
SIMSTAT provides most of the simple statistical tests
and procedures that you will ever need. One exception is nonlinear
regression, which can lead to non-unique results and which should
therefore be performed (except for ARIMA and time series)
using one of the dedicated Simfit nonlinear fitting programs.
Many of the procedures provided
by SIMSTAT are also available for experienced users but
with more help and advice, and sometimes with additional options,
as dedicated individual programs. For example, program LINFIT
can be used for robust regression as well as for linear
and multilinear regression. You can prepare your data in files
like the test files supplied with the Simfit package, but many
users prefer to keep their spreadsheet program open at the same time
as Simfit and simply transfer data as required using the clipboard.
Whenever a data file is requested by Simfit the clipboard is inspected
and, if the clipboard contains a data table, the Paste button becomes
activated so you can paste the data directly into Simfit. You can
inspect the clipboard data at any time from the main Simfit [View]
menu.
Preparing data
The Simfit package has a special set of editors to allow you
to create or edit data files for statistical analysis.
If you use these editors your data
will always be correctly formatted and will never be lost by
overwriting. These editors have numerous special features and
powerful options for manipulating data and graphs, and you are
recommended to use them. Often, however, users already have their
data in the form of tables in spreadsheet programs, so Simfit
can analyse tables pasted from the clipboard or create
correctly formatted files from clipboard data.
Users of Microsoft Excel can create
files in the Simfit format directly from Excel by using the Simfit
macros called Simfit?.xls, e.g. Simfit4.xls. These are loaded
into Excel and can be activated to create files in the Simfit
format from data columns that you select by highlighting.
To analyse a single sample you need a column (i.e. vector) of data
in the correct format. This can be typed in from the terminal
or pasted in from the clipboard, but the best way is to make a vector
file either by using program MAKMAT or by pasting clipboard data
into program MAKSIM.
To analyse two or more samples simply requires a set of such files
but, for most procedures involving several vectors, it is much
better to create a library file using program MAKLIB, so that a
whole group of files can be selected by reading in just one
library file.
Where all the columns are of the same length, as in contingency table
analysis or when exploring multiple correlations, it is possible to
prepare a matrix file by using program MAKMAT, or by pasting a selected
table from the clipboard directly into SIMSTAT, or into a file using
program MAKSIM. You can, however, use a library file for most
matrix calculations, except for contingency table analysis, the
Fisher exact test and linear algebra calculations.
Test files
The Simfit package has a set of test files so you can try with
data known to be appropriate before trying with your own data.
These are referred to when the individual analytical procedures
are described. To analyse test data just select the Demo button
when data input is requested.
Missing data
Files supplied to SIMSTAT can have no missing values. These must
be replaced by substituted values, using your own scheme or one of the
methods provided by the Simfit package.
SIMSTAT options
- Data exploration:
choose this option when you want to investigate vectors or matrices
for means, medians, variances, appearance when plotted, etc.
- Standard statistical tests:
choose this option when you want to perform tests on data
such as Kolomorov-Smrirnov tests, Mann-Whitney U tests, t tests,
chi-square tests, tests for a normal distribution, etc.
- Analysis of variance:
choose this option when you want to investigate sets of continuous
variables.
- Analysis of proportions:
choose this option when you want to investigate sets of
discrete variables.
- Multivariate statistics:
choose this option for correlation analysis, to create dendrograms,
calculate principal components, or compare groups using MANOVA, or
canonical variates.
- Regression and calibration:
choose this option if you want to fit a line to data,
prepare a standard curve for calibration, fit a polynomial,
do multilinear regression, or fit a dose response curve for EC50
or LD50, etc.
- Generalized linear models:
choose this option for logistic regression, binary logistic regression,
or survival analysis with covariates using Exponential, Weibull or Cox
models, etc.
- Time series and survival times:
choose this option to analyse data obtained as a sequence of measurements
at equal time increments and you want to do data smoothing, estimate
autocorrelations or
or fit autoregressive integrated moving average models (ARIMA).
Survival analysis of data with no covariates is also available.
- Statistical calculations:
choose this option when you want to do calculations on
exploring the variation in power as a function of sample size
when preparing protocols for clinical trials,
estimating parameter confidence limits, or generating random numbers.
- Numerical analysis:
choose this option when you want to do such procedures as
finding zeros of polynomials, eigenvalues, determinants, inverses,
singular value decompositions of matrices, matrix factorisations or
solving linear equations,
Back to Help Menu or End Help