In Reply To:
In Reply To:
In Reply To:
Just to clarify, you think that observer expectancy and subject expectancy effects are far-fetched? They do sound far-fetched until you realise how powerful and well-supported by a massive body of literature they are. As both subject and observer in his study, the potential for Tom's expectancy to bias the results is enormous. If you design a study like this, you may find what you expect to find, but it doesn't make it true.
When your kid's forehead feels hot, do you hand the thermometer to 8 different people to take her temperature? If not, how do you know that your observer expectancy hasn't rendered any rule of thumb that says, "my kid's temperature is 40C and needs to be reduced" invalid?
While this seems like a straight forward example, taking a temperature, while seeming simple, is filled with potential error. Most of the errors would lead to a "too low" reading but some can result in "too high". Was the thermometer shook down properly after the last temp for instance. Or, did mom just give the kid some hot soup or tea and you come in to take the temperature. The kid may try to manipulate the temperature if he knows the right number will keep him out of school or keep him out of the ice bath.
It is unlikely that observer bias would play a role in taking a temperature, but it certainly does in taking a simple measurement like blood pressure. People keep taking it until they get the number they want, then take that one.
All measurements are prone to error and due care must be used in both obtaining the measurement and in interpreting it (if the kids temp is measured as 40C but he is happily playing video games, does one necessarily believe it?).
I think the thermometer example is a good one. We're not too concerned whether the true temperature is 39.9 or 40.1; what we're typically interested in is whether the measured temperature is greater than or less than the rule of thumb threshold for a fever. We don't give that thermometer to 8 other people and ask them to repeat the reading because we already have an idea of what the precision is for this device, and how to relate the measurement to other observations about the patient's condition. We know those things because we've used thermometers many, many times in the past and have learned how to use it, to interpret it, and how to recognize that a reading of 30C or 50C is way outside the usual precision. Not knowing whether the true temperature is 39.9 or 40.1 does not invalidate the rule of thumb "higher than 38 means a fever." It's a rule of thumb.
In this case, Tom has used this method many, many times in the past and has learned how to do it, to interpret the results, and to recognize when the results are spoiled (as he did for the final run). The measured difference in CdA is an order of magnitude greater than his usual precision.