Example 2

The second example analyzes reviewing behavior of individual reviewers by aggregating reviewer file and then adding in reviewer characteristics and author information from sociologist file and aggregated manuscript information.

1. Do reviewers favor particular types of authors (For example do viewers with certain demographic or institutional characteristics more generous to authors with similar characteristics)?
2. What types of reviewers are most generous or stingy with acceptances or R&Rs?

SPSS syntax statement is provided below.  Non-SPSS users should look at the logic, which can then be replicated with the appropriate SAS of STATA commands.

Unit of Analysis: Reviewers with Summary Information on All Reviews Prepared

* Begin with review file 

GET

FILE='Location and name of manuscript file'. 

/* initialize outcome variables to 0 and set to 1 for each instance of an outcome

do repeat outcome = Accept CondAccept NA RandR Reject .

compute outcome = 0.

end repeat.

if (recommend eq 'Accept' ) accept = 1.

if (recommend eq 'Cond Accept' ) condaccept = 1.

if (recommend eq 'N/A' ) NA = 1.

if (recommend eq 'R & R') RandR = 1.

if (recommend eq 'Reject') reject = 1.

frequencies variables = Accept CondAccept NA RandR Reject .

/* aggregate command yields total N of reviews as well as of each outcome

DATASET DECLARE AGG_Reviews.

SORT CASES BY PersonID.

AGGREGATE

  /OUTFILE='AGG_Reviews'

  /PRESORTED

  /BREAK=PersonID

  /Accept_sum=SUM(Accept)

  /CondAccept_sum=SUM(CondAccept)

  /NA_sum=SUM(NA)

  /RandR_sum=SUM(RandR)

  /Reject_sum=SUM(Reject)

  /N_of_reviews=N.

/* calculate percents for each outcome

compute perAccept = Accept_sum / N_of_reviews.

compute perCondAccept = CondAccept_sum / N_of_reviews.

compute perNA = NA_sum / N_of_reviews.

compute perRandR = RandR_sum / N_of_reviews.

compute perReject = Reject_sum / N_of_reviews.

descriptives variables = perAccept perCondAccept perNA perRandR perReject N_of_reviews.

/* match in individual level characteristics for each reviewer

MATCH FILES /FILE= *

  /TABLE= 'Location and name of manuscript file'. 

  /BY personID .

EXECUTE.

/* transform alphanumberic gender variable into nominal numeric

compute gender_num =3.

if (gender eq 'male') gender_num = 1.

if (gender eq 'female') gender_num = 2.

value labels gender_num 1 'Male'  2 'Female'  3  'Other'.

frequencies variables = gender_num gender.

/* temporarily select only men and women and run t-test - men do more reviews and accept a greater percent

temporary.

select if (gender_num le 2).

T-TEST GROUPS=gender_num(1 2)

  /MISSING=ANALYSIS

  /VARIABLES=N_of_reviews perAccept, perCondAccept, perRandR, perReject

  /CRITERIA=CI(.95).

From this example, in Tables 4 and 5, we can see 62.1% of all those who reviewed for ASR during this time period were male, but also that on average male reviews did significantly more reviews (5.23) than female reviewers (4.26).  However, there was little difference in the decisions they made: the percentage acceptance rate was significantly higher for men than women (p=.05), but there was no significant difference according to gender for any other outcome.

Table 4

Table 5

There are more older reviewers than younger reviewers and the older viewers were significantly more likely to recommend accepting a manuscript and significantly less likely to outright reject a manuscript than younger reviewers. (Table 6)

Table 6