Discussing research – day 5: language for discussing factors affecting data collection

Preview

When writing about data collection, it’s often necessary to discuss factors that may have affected your data.

Nouns

A bias – meaning: when a situation or someone’s opinion moves unfairly towards one side or result

These adjectives are often used with this noun: slight, potential, overt
When bias was not intended it is ‘unconscious bias’
When bias is in the system/organisation it is ‘inherent bias’, ‘institutional bias’ or ‘systematic bias’
This noun is often used with these prepositions: against, towards

The inherent bias of the institution against students from state schools is visible in its poorly developed outreach programme.

Impartiality – meaning: when you are unconnected to a particular person, group, or viewpoint

We can use the phrases ‘a lack of impartiality’ or ‘an inability to remain impartial’ to describe a situation where someone/something is not impartial (adjective).

The juror excused himself from the trial, citing and inability to remain impartial as his reason for leaving.

A constraint – meaning: something that stops your form doing what you want to do

These adjectives are often used with this noun: key, significant, severe, tight
These verbs are often used with this noun: impose, face, overcome, remove

In order to improve economic development in the area, the government has removed constraints on the planning and construction of new commercial buildings.

A limitation – meaning: a fact, rule, situation that limits something OR a disadvantage

These adjectives are often used with this noun: basic, fundamental, obvious, severe, serious, major, significant,
These verbs are often used with this noun, particularly when the limitation refers to a rule: impose, place, remove

The researchers discussed the limitations of their work.

A systematic error – meaning: a recurring error that is caused by inaccuracy, inadequate methods or failings connected with equipment

Due to a systematic error in survey distribution, none of the respondents were over 35.

Misclassification – meaning: when something is put into the wrong category or group

By expanding the number of responses available, the survey creators can reduce the risk of misclassification.

Scope – meaning: the range of things that something deals with

These verbs are often used with this noun: narrow, reduce, restrict, minimise, broaden, extend, increase, widen,
These nouns are often used after ‘the scope of’: study, work, research, project, analysis, investigation

There were a variety of factors that were responsible for restricting the scope of our investigation.

Noise – meaning: information or parts of data that are unimportant and that make the data unclear

It is important to consider the effects of data noise on our findings.

Phrases

It is not error free OR it is not without error – meaning: there are mistakes

The process was not without error so alterations have been made to improve future accuracy.

To (not) be fully representative – meaning: to (not) be typical of particular group or to by a typical example of something

These adverbs are often used with the adjective ‘representative’: hardly, fairly, sufficiently reasonably, fully, highly
These nouns are often used with ‘representative’: cross-section, sample, selection, view, example

The respondents were not fully representative of the local community.

Verbs

To compromise – meaning: to put someone or something at risk

Negative press surrounding the CEO compromised the success of the product launch.

This verb is often used in the passive form.

The study was compromised by ineffective equipment.

To expose – meaning: to place someone/something near something harmful

These adverbs are often used with this verb: temporarily, repeatedly, constantly, directly, indirectly
This verb is usually followed by the preposition ‘to’ before an object
This verb is often used in the passive form

Due to misclassification in the laboratory, the samples were temporarily exposed to harmful gases.

To deviate – meaning: to do something differently to the way that was agreed

These adverbs are often used with this verb: slightly, significantly, substantially
This verb is usually followed by the preposition ‘from’

The contractors deviated from the original plan.

The extract below contains examples of some of the language covered in this lesson.

Epidemic dreams: dreaming about health during the COVID-19 pandemic

Conclusion

This study has three main limitations that call for future work. The first comes from the potential biases introduced by the data collection process. We collected dream reports specifically concerning COVID-19 (i.e. the users were primed to talk about it), while tweets were written by users who freely decided to vent about COVID-19 (i.e. they self-selected for it). That is why we used a rank-based method to compare the health mentions in the two datasets.

The second limitation has to do with our method for extracting mentions of medical conditions from text. Although MedDL is a state-of-the-art tool with top-class accuracy, its output is not error-free. Because MedDL was trained on social media data only, misclassifications could be more frequent when applied to the dream reports dataset. Our qualitative analysis did not produce evidence for any systematic error that would compromise our results. However, future work could collect additional training data specific to dream reports.

The third limitation has to do with the quality and scope of our two datasets. Our Twitter dataset, albeit large, is not fully representative of the general population. Studies on Twitter are exposed to issues of data noise [43,44], representativeness [45] and self-presentation biases [46]. In the USA, the country in which Twitter has highest penetration rate, socio-demographic characteristics deviate from the general population: Twitter users are much younger, with a higher level of formal education, and are more likely to support the Democratic Party [32]. Our collection of dream reports has limitations too. We gathered the reports through a web survey without imposing limits or constraints on the input text, which could introduce some noise in the data.

This extract is taken from: Šćepanović Sanja, Aiello Luca Maria, Barrett Deirdre, Quercia Daniele, 2022 Epidemic dreams: dreaming about health during the COVID-19 pandemic R. Soc. Open, sci. 9211080211080 http://doi.org/10.1098/rsos.211080

There is no multiple-choice quiz with today’s article.

Lesson tags: discussing research, English for scientists, language for data collection