rempsyc / busara_dashboard

The Missing Majority in Behavioural Science Dashboard
https://remi-theriault.com/dashboards/missing_majority
1 stars 0 forks source link

Strange results in PubMed Query as a function of the number of journals #26

Closed rempsyc closed 8 months ago

rempsyc commented 8 months ago

This query returns zero result:

Developmental psychology [Journal] OR Journal of personality and social psychology [Journal] OR Journal of abnormal psychology [Journal] OR Journal of family psychology : JFP : journal of the Division of Family Psychology of the American Psychological Association (Division 43) [Journal] OR Health psychology : official journal of the Division of Health Psychology, American Psychological Association [Journal] OR Journal of educational psychology [Journal] OR Journal of experimental social psychology [Journal] OR Collabra. Psychology [Journal] OR Journal of experimental psychology. General [Journal] OR The Journal of applied psychology [Journal] OR Psychological methods [Journal] OR Advances in methods and practices in psychological science [Journal] OR Psychological science [Journal] OR Child development [Journal] OR Developmental science [Journal] OR Personality and social psychology bulletin [Journal] OR Nature human behaviour [Journal] OR Journal Of economic psychology [Journal] OR Journal of experimental and behavioral economics [Journal] OR Experimental economics [Journal] OR Journal of development economics [Journal] OR World development [Journal] OR Quarterly journal of economics [Journal] OR Econometrica : journal of the Econometric Society [Journal] OR Behavioral public policy [Journal] OR African development review [Journal] OR African journal of agricultural and resource economics [Journal] OR African journal of economic and management studies [Journal] OR African journal of economic policy [Journal] OR American economic journal. Applied economics [Journal] OR American economic journal. Economic policy [Journal] OR American economic journal. Macroeconomics [Journal] OR The American economic review [Journal] OR American economic review. Insights [Journal] OR Economic development and cultural change [Journal] OR Economic journal (London, England) [Journal] OR Journal of African economies [Journal] OR Journal of African development [Journal] OR Journal of human development and capabilities [Journal] OR Journal of development effectiveness [Journal] OR Journal of development studies [Journal] OR Journal of economic growth (Boston, Mass.) [Journal] OR Journal of labor economics [Journal] OR Journal of political economy [Journal] OR Journal of public economics [Journal] OR Review of African political economy [Journal] OR Review of development economics [Journal] OR The Review of economic studies [Journal] OR Review of international political economy [Journal] OR South African journal of economics [Journal] OR The World Bank economic review [Journal] OR World bank Research observer [Journal] OR World development perspectives [Journal] AND ('2000/01/01 [Date - Publication] : '2000/12/31 [Date - Publication])

This one 68 results:

Developmental psychology [Journal] OR World development perspectives [Journal] AND ('2000/01/01 [Date - Publication] : '2000/12/31 [Date - Publication])

It is not clear why using more choices of journals (using the OR operator) would lead to fewer results than filtering for fewer journals. This should be investigated because it probably impacts our data.

rempsyc commented 8 months ago

It could be that the query is too long, which would require us to split queries into several calls...

rempsyc commented 8 months ago

More investigation:

No results:

Developmental psychology [Journal] OR Journal of economic growth (Boston, Mass.) [Journal] AND ('2000/01/01 [Date - Publication] : '2000/12/31 [Date - Publication])

68 results:

Developmental psychology [Journal] [AND ('2000/01/01 [Date - Publication] : '2000/12/31 [Date - Publication])

Also 68 results after changing the order:

Journal of economic growth (Boston, Mass.) [Journal] OR Developmental psychology [Journal] AND ('2000/01/01 [Date - Publication] : '2000/12/31 [Date - Publication])

So seems the problem originates from the year criteria applying only to the last result... There is this bit of information online:

Using Parentheses to Create an Order of Operations

In PubMed there are no separate search statements to indicate the order of performance for the Boolean operations. Instead, parentheses are the indicators for which operations are performed in what order, e.g., (a OR b) & (c OR d) means that the OR operations inside the parentheses will be done before the AND operation between the two sets of terms. If there are no parentheses used in a search formulation, then the operations will be processed left to right.

But using parentheses in our case does not solve the problem:

No results:

(Developmental psychology [Journal] OR Journal of economic growth (Boston, Mass.) [Journal]) AND ('2000/01/01 [Date - Publication] : '2000/12/31 [Date - Publication])

68 results:

(Journal of economic growth (Boston, Mass.) [Journal] OR Developmental psychology [Journal]) AND ('2000/01/01 [Date - Publication] : '2000/12/31 [Date - Publication])

So the priority order is not respected. I wonder if this has to do with the parenthesis already in use for the economic journal (Boston, Mass.). It seems like it, yes:

68 results:

(Developmental psychology [Journal] OR Journal of economic growth [Journal]) AND ('2000/01/01 [Date - Publication] : '2000/12/31 [Date - Publication])
rempsyc commented 8 months ago

I posted a question on stackoverflow here: https://stackoverflow.com/questions/78151974/pubmed-query-priority-of-operations-using-parenthesis-leads-to-zero-results

rempsyc commented 8 months ago

A workaround to try later would be to use parentheses for all journal names, perhaps that will solve the issue.

rempsyc commented 8 months ago

The fix seems to be to use quotes around journal names (especially for journal names with parentheses). However, one source explains,

Not all phrases (string of words you enclose in double quotes) can be found in PubMed due to how PubMed indexes phrases. The easiest solution to this error is remove the double quotes, however this can lead to unintended results if you are not careful.

rempsyc commented 8 months ago

The following query for example,

"Journal of experimental and behavioral economics" [Journal]

Provides zero results, with the following warning:

The following term was ignored: Journal of experimental and behavioral economics

But it is not clear why this term is ignored!?

If I do it without quotes,

Journal of experimental and behavioral economics [Journal]

Then we get

The following terms were ignored: Journal, of, experimental, and, behavioral, economics
rempsyc commented 8 months ago

It seems SEVERAL of the original psych journals are lost that we added more journals.

Before:

image

After:

image
rempsyc commented 8 months ago

I think the final solution was to use double quotes instead of single quotes in the PubMed query... quite the relief after all this troubleshooting!