# This file contains all of the commands used to analyze the inter1
# experiment.  It is a good idea to store a record of every analysis command
# that you run.  Anything in this file that is not a command should be
# commented out, like this text is, so that you can simply run this file as an
# executable script to run all of the analyses.  To make this file executable,
# run this command:
# > chmod a+x analyze

# This first command checks for bad subjects, using only the data from the
# filler items and other experiments.  You may want to replace any subjects
# with very low question answering performance or very slow reading times.  To
# replace a subject, move his or her .dat file to a backup location and re-run
# a new subject using the same subject number.  The reason for using the same
# subject number is that the same items will be chosen, so your conditions
# should remain balanced.  In this example, we won't bother to discard the bad
# subjects.
lingalyzer inter1.cnd -f '![in $EXPT "practice inter1"]' -p ../Results

# This command collects the data from the .dat files, computes various
# statistics, and generates a .rtm and a .qst summary file, which are
# compressed:
lingalyzer inter1.cnd -c ../Results

# This checks for any words that were not included in any conditions.  Any
# results will be printed to standard output.  If there are results, you may
# need to fix the region definitions in the .cnd file:
lingalyzer inter1.cnd -i '$EXPT $ITEM $COND' -d '$WNUM $WORD' -f '[eq $EXPT inter1] && $RNUM==0' -e -r -


############################ READING TIME GRAPHS ##############################

# This compiles the average raw reading times for each condition and region
# without trimming any data and graphs the results using lingrapher.  It will
# produce an encapsulated postscript file called raw-notr.eps which can be
# viewed with gv, if available.  It will also produce an xmgrace file called
# raw-notr.agr.  This can be opened using xmgrace if you want to modify the
# graph or generate images in other formats, such as jpg.
lingalyzer inter1.cnd -i '$EXPT $COND $RNUM' -d '$RWRT' -f '[eq $EXPT inter1]' -r raw-notr
lingrapher inter1.cnd raw-notr.avg raw-notr raw.lpm

# Generates a raw reading time graph using trimmed data:
lingalyzer inter1.cnd -i '$EXPT $COND $RNUM' -d '$RWRT' -f '[eq $EXPT inter1] && $RWZS < 4' -r raw-trim
lingrapher inter1.cnd raw-trim.avg raw-trim raw.lpm

# Generates a residual reading time graph using un-trimmed data:
lingalyzer inter1.cnd -i '$EXPT $COND $RNUM' -d '$RSRT' -f '[eq $EXPT inter1]' -r res-notr
lingrapher inter1.cnd res-notr.avg res-notr res.lpm
 
# Generates a residual reading time graph using trimmed data:
lingalyzer inter1.cnd -i '$EXPT $COND $RNUM' -d '$RSRT' -f '[eq $EXPT inter1] && $RSZS < 4' -r res-trim
lingrapher inter1.cnd res-trim.avg res-trim res.lpm


###################### STATISTICS ON THE WHOLE SENTENCE #######################

# To get average trimmed raw RTs on the whole sentence (produces raw-whole.avg):
lingalyzer inter1.cnd -i '$COND' -d '$RWRT' -f '[eq $EXPT inter1] && $RWZS < 4' -r raw-whole

# To compute an ANOVA by subjects on the average trimmed raw RTs on the whole
# sentence (produces raw-whole-s.anv and raw-whole-s.anova):
lingalyzer inter1.cnd -i '$SUBJ $ANOV' -d '$RWRT' -a 'subject type length raw' -f '[eq $EXPT inter1] && $RWZS < 4' -r raw-whole-s

# To compute pairwise T-tests for every pair of conditions on the average
# trimmed raw RTs on the whole sentence (produces raw-whole-s.pwa):
lingalyzer inter1.cnd -i '$SUBJ $COND' -d '$RWRT' -t 'subject condition raw' -f '[eq $EXPT inter1] && $RWZS < 4' -r raw-whole-s

# Raw RT ANOVAs and pairwise T-tests by items (produces raw-whole-i.anv,
# raw-whole-i.anova, and raw-whole-i.pwa):
lingalyzer inter1.cnd -i '$ITEM $ANOV' -d '$RWRT' -a 'item type length raw' -f '[eq $EXPT inter1] && $RWZS < 4' -r raw-whole-i
lingalyzer inter1.cnd -i '$ITEM $COND' -d '$RWRT' -t 'item condition raw' -f '[eq $EXPT inter1] && $RWZS < 4' -r raw-whole-i

# To get average trimmed residual RTs on the whole sentence (produces res-whole.avg):
lingalyzer inter1.cnd -i '$COND' -d '$RSRT' -f '[eq $EXPT inter1] && $RSZS < 4' -r res-whole

# Residual RT ANOVAs and pairwise T-tests by subjects (produces
# res-whole-s.anv, res-whole-s.anova, res-whole-s.pwa):
lingalyzer inter1.cnd -i '$SUBJ $ANOV' -d '$RSRT' -a 'subject type length resid' -f '[eq $EXPT inter1] && $RSZS < 4' -r res-whole-s
lingalyzer inter1.cnd -i '$SUBJ $COND' -d '$RSRT' -t 'subject condition resid' -f '[eq $EXPT inter1] && $RSZS < 4' -r res-whole-s

# Residual RT ANOVAs and pairwise T-tests by items (produces
# res-whole-i.anv, res-whole-i.anova, res-whole-i.pwa):
lingalyzer inter1.cnd -i '$ITEM $ANOV' -d '$RSRT' -a 'item type length resid' -f '[eq $EXPT inter1] && $RSZS < 4' -r res-whole-i
lingalyzer inter1.cnd -i '$ITEM $COND' -d '$RSRT' -t 'item condition resid' -f '[eq $EXPT inter1] && $RSZS < 4' -r res-whole-i


########################### STATISTICS ON REGION 6 ############################

# Averages, ANOVAS, T-tests on raw data:
lingalyzer inter1.cnd -i '$COND' -d '$RWRT' -f '[eq $EXPT inter1] && $RWZS < 4 && $RNUM==6' -r raw-r6
lingalyzer inter1.cnd -i '$SUBJ $ANOV' -d '$RWRT' -a 'subject type length raw' -f '[eq $EXPT inter1] && $RWZS < 4 && $RNUM==6' -r raw-r6-s
lingalyzer inter1.cnd -i '$SUBJ $COND' -d '$RWRT' -t 'subject condition raw' -f '[eq $EXPT inter1] && $RWZS < 4 && $RNUM==6' -r raw-r6-s
lingalyzer inter1.cnd -i '$ITEM $ANOV' -d '$RWRT' -a 'item type length raw' -f '[eq $EXPT inter1] && $RWZS < 4 && $RNUM==6' -r raw-r6-i
lingalyzer inter1.cnd -i '$ITEM $COND' -d '$RWRT' -t 'item condition raw' -f '[eq $EXPT inter1] && $RWZS < 4 && $RNUM==6' -r raw-r6-i

# Averages, ANOVAS, T-tests on residual data:
lingalyzer inter1.cnd -i '$COND' -d '$RSRT' -f '[eq $EXPT inter1] && $RSZS < 4 && $RNUM==6' -r res-r6
lingalyzer inter1.cnd -i '$SUBJ $ANOV' -d '$RSRT' -a 'subject type length resid' -f '[eq $EXPT inter1] && $RSZS < 4 && $RNUM==6' -r res-r6-s
lingalyzer inter1.cnd -i '$SUBJ $COND' -d '$RSRT' -t 'subject condition resid' -f '[eq $EXPT inter1] && $RSZS < 4 && $RNUM==6' -r res-r6-s
lingalyzer inter1.cnd -i '$ITEM $ANOV' -d '$RSRT' -a 'item type length resid' -f '[eq $EXPT inter1] && $RSZS < 4 && $RNUM==6' -r res-r6-i
lingalyzer inter1.cnd -i '$ITEM $COND' -d '$RSRT' -t 'item condition resid' -f '[eq $EXPT inter1] && $RSZS < 4 && $RNUM==6' -r res-r6-i

############################# QUESTION ANSWERING ##############################

# Average question answering percent correct by condition:
lingalyzer inter1.cnd -i '$EXPT $COND' -d '$QCRC' -q qst

# ANOVA and pairwise T-Tests by subjects:
lingalyzer inter1.cnd -i '$SUBJ $ANOV' -d '$QCRC' -a 'subject type length qcrc' -f '[eq $EXPT inter1]' -q qst-s
lingalyzer inter1.cnd -i '$SUBJ $COND' -d '$QCRC' -t 'subject condition qcrc' -f '[eq $EXPT inter1]' -q qst-s

# ANOVA and pairwise T-Tests by items:
lingalyzer inter1.cnd -i '$ITEM $ANOV' -d '$QCRC' -a 'item type length qcrc' -f '[eq $EXPT inter1]' -q qst-i
lingalyzer inter1.cnd -i '$ITEM $COND' -d '$QCRC' -t 'item condition qcrc' -f '[eq $EXPT inter1]' -q qst-i

# Average question answering times:
lingalyzer inter1.cnd -i '$EXPT $COND' -d '$RWRT' -q qst-rt
lingalyzer inter1.cnd -i '$SUBJ $ANOV' -d '$RWRT' -a 'subject type length qcrc' -f '[eq $EXPT inter1]' -q qst-rt-s
lingalyzer inter1.cnd -i '$SUBJ $COND' -d '$RWRT' -t 'subject condition qcrc' -f '[eq $EXPT inter1]' -q qst-rt-s

############################## REGRESSION ANALYSIS ############################

# Although it is not necessarily appropriate for this experiment, this is an
# example of how to perform a reading time regression analysis.  This may be
# useful if you, for example, have three conditions that vary in a single
# dimension, such as short, medium, long or easy, medium hard.

# The Difficulty array has been defined in the inter1.cnd to hold an integer
# value for each of the conditions, reflecting their expected difficulty,
# according to our favorite theory.  We will compute a regression of the
# residual reading time against these values to see if there is a significant
# trend towards longer reading times with more difficult items.

# The independent variable is the difficulty level and the dependent is the
# residual reading time.  The -e option is used to extract the individual
# datapoints and - is used as the output filename, so the output is sent to
# standard out, which is piped to the regress command.

lingalyzer inter1.cnd -i '$Difficulty($COND)' -d '$RSRT' -f '[eq $EXPT inter1] && $RSZS < 4 && $RNUM==6' -e -r - | regress > res-r6.regress